repository stringclasses 156 values | issue title stringlengths 1 1.01k ⌀ | labels stringclasses 8 values | body stringlengths 1 270k ⌀ |
|---|---|---|---|
openzipkinzipkin | two separate bugfixe | Bug | 1 bugfix gototrace not trigger on other page than default create a malformed url 2 webpack devproxy only target default page locally and get resource from upstream api if you attach an external dev server the external zipkin server can now be configure with an environment variable |
openzipkinzipkin | ui fail to render trace without endpoint | Bug | the thrift l276 schema state that endpoint be optional however when render it appear that for some core annotation the endpoint be require and override the value see this l132 otherwise the page fail to render |
openzipkinzipkin | zipkin web trace display error | Question | the trace show be out of order use golang image theoretically it should be like this use c image json datum generate by the same two language image and sometimes on the first page after click find trace trace time display will be more than 100 image if you click it you can see image I retrieve all its span of trace i d and parentid be correct my system be work mode 1 custom span datum transmit from the server be convert to zipkin span of json datum 2 then send json datum via http post method to query server I do not know zipkin system for json datum be send to the query server must be whether a certain order trace of the same order and there may be display correctly there be likely to fail just like the above two error currently the test environment be to use docker deployment |
openzipkinzipkin | fix ui display of json value | Bug | before json in a binary annotation value result in the ui display object object this guard and stringifie object while leave primitive alone fix 1113 |
openzipkinzipkin | display service name when show address in the span detail element | Enhancement | example screenshot image why when a request come from or go to uninstrumented service the instrumented part may log can sa annotation that include not only the peer ip address but also the service name however the current ui while it have a special handing for ca sa annotation do not display the service name |
openzipkinzipkin | ui do not display binary annotation value if it contain a json string | Enhancement | our instrumentation emit some binary annotation as json string when return from query service it look all properly escape key haproxy logmessage value msg tw 0 tt 8625 tr 0 tq 8624 tc 1 backend but in the span popup it be display like this haproxy logmessage object object cc eirslett |
openzipkinzipkin | zipkin ui text roll off the rhs | Enhancement | would be nice if we could do something to prevent text roll off the screen image json traceid daeea3ed1a5c9002 name query i d daeea3ed1a5c9002 timestamp 1462244752792000 duration 406000 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244752792000 value sr endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753198000 value ss binaryannotation key client value 127 0 0 1 endpoint servicename c test cluster localhost ipv4 172 18 0 1 key coordinator value localhost 127 0 0 1 endpoint servicename c test cluster localhost ipv4 172 18 0 1 key request value execute cql3 query endpoint servicename c test cluster localhost ipv4 172 18 0 1 key sessionid value f2263010 10db 11e6 b320 75631f677eb6 endpoint servicename c test cluster localhost ipv4 172 18 0 1 key start at value 2016 05 03t03 05 52 875z endpoint servicename c test cluster localhost ipv4 172 18 0 1 traceid daeea3ed1a5c9002 name parse select from system size estimate sharedpool worker 1 i d 3a7bfb16f185ae76 parentid daeea3ed1a5c9002 timestamp 1462244752962000 duration 0 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244752962000 value cr endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244752962000 value cs binaryannotation traceid daeea3ed1a5c9002 name prepare statement sharedpool worker 1 i d e0f9fac3cb371a53 parentid daeea3ed1a5c9002 timestamp 1462244753176000 duration 0 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753176000 value cr endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753176000 value cs binaryannotation traceid daeea3ed1a5c9002 name computing range to query sharedpool worker 1 i d c2d493893e1c5de1 parentid daeea3ed1a5c9002 timestamp 1462244753178000 duration 1000 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753178000 value cs endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753179000 value cr binaryannotation traceid daeea3ed1a5c9002 name execute seq scan across 3 sstable for min 9223372036854775808 min 9223372036854775808 sharedpool worker 2 i d 3b075942792a910b parentid daeea3ed1a5c9002 timestamp 1462244753181000 duration 3000 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753181000 value cs endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753184000 value cr binaryannotation traceid daeea3ed1a5c9002 name submit range request on 1 range with a concurrency of 1 12 15 row per range expect sharedpool worker 1 i d b6d971ed42a67aad parentid daeea3ed1a5c9002 timestamp 1462244753181000 duration 0 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753181000 value cr endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753181000 value cs binaryannotation traceid daeea3ed1a5c9002 name submit 1 concurrent range request sharedpool worker 1 i d 3192b848f6fb8540 parentid daeea3ed1a5c9002 timestamp 1462244753183000 duration 7000 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753183000 value cs endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753190000 value cr binaryannotation traceid daeea3ed1a5c9002 name read 16 live and 0 tombstone cell sharedpool worker 2 i d f3e1483ed4be9c4c parentid daeea3ed1a5c9002 timestamp 1462244753188000 duration 6000 annotation endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753188000 value cs endpoint servicename c test cluster localhost ipv4 172 18 0 1 timestamp 1462244753194000 value cr binaryannotation |
openzipkinzipkin | make servicename an option string in queryrequest | Enhancement | issue zipkin query currently have no way to request trace across service make it difficult to show any sort of int service overview ui propose solution change servicename from be require to be optional I m new to cql so feedback on repository java change would be appreciate |
openzipkinzipkin | zipkin ui could provide well feedback on slow response | Enhancement | when an api query be slow it could look misallocated in chrome debug for example the below slow call to service wasn t cause by api v1 service rather api v1 trace which take 200second to return thank to yzhang8015 for report |
openzipkinzipkin | can not find trace in ui | Question | I input traceid in ui web curl I response be 404 how can I input traceid find trace info |
openzipkinzipkin | add save a trace functionality | Enhancement | it d be really nice if individual trace could be tag to not age out of cassandra through the ui |
openzipkinzipkin | not all the span render in the ui | Bug | give the following trace where I see a lot of servicea there be 10 I only see two of they render in the ui which seem very weird the ui in fact seem to keep space time to render they but do not render anything traceid bd8283279c35f5e1 name get i d 54f3362c23fd3423 parentid bd8283279c35f5e1 timestamp 1460956558786749 duration 19353 annotation timestamp 1460956558786761 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956558806099 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d b9a14937a2cfa97a parentid bd8283279c35f5e1 timestamp 1460956558820804 duration 21045 annotation timestamp 1460956558820813 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956558841846 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d e2218aab4a57245a parentid bd8283279c35f5e1 timestamp 1460956558892976 duration 208907 annotation timestamp 1460956558893007 value sr endpoint servicename ui app ipv4 10 225 145 5 port 3000 timestamp 1460956559101880 value ss endpoint servicename ui app ipv4 10 225 145 5 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d 4c932d3d10182aa9 parentid e2218aab4a57245a timestamp 1460956558897019 duration 135474 annotation timestamp 1460956558897037 value cs endpoint servicename ui app ipv4 10 225 145 5 port 3000 timestamp 1460956559032491 value cr endpoint servicename ui app ipv4 10 225 145 5 port 3000 binaryannotation key http status value 200 endpoint servicename ui app ipv4 10 225 145 5 port 3000 key http uri value v1 privilege show endpoint servicename ui app ipv4 10 225 145 5 port 3000 key sa value 1 endpoint servicename serviceb ipv4 54 86 129 159 port 443 debug false traceid bd8283279c35f5e1 name get i d 9c85a02c27fe70fd parentid 4c932d3d10182aa9 timestamp 1460956558899306 duration 129897 annotation timestamp 1460956558899312 value cs endpoint servicename ui app ipv4 10 225 145 5 port 3000 timestamp 1460956559029196 value cr endpoint servicename ui app ipv4 10 225 145 5 port 3000 binaryannotation key http status value 200 endpoint servicename ui app ipv4 10 225 145 5 port 3000 key http uri value v1 privilege show endpoint servicename ui app ipv4 10 225 145 5 port 3000 key sa value 1 endpoint servicename serviceb ipv4 54 86 129 159 port 443 debug false traceid bd8283279c35f5e1 name get i d 08a8444fc69d5ef3 parentid e2218aab4a57245a timestamp 1460956559035581 duration 57335 annotation timestamp 1460956559035590 value cs endpoint servicename ui app ipv4 10 225 145 5 port 3000 timestamp 1460956559092913 value cr endpoint servicename ui app ipv4 10 225 145 5 port 3000 binaryannotation key http status value 200 endpoint servicename ui app ipv4 10 225 145 5 port 3000 key http uri value v1 study environment assignment endpoint servicename ui app ipv4 10 225 145 5 port 3000 key sa value 1 endpoint servicename servicea ipv4 107 23 29 208 port 443 debug false traceid bd8283279c35f5e1 name get i d d785650bad29dd26 parentid 08a8444fc69d5ef3 timestamp 1460956559036652 duration 55860 annotation timestamp 1460956559036658 value cs endpoint servicename ui app ipv4 10 225 145 5 port 3000 timestamp 1460956559073798 value sr endpoint servicename servicea ipv4 10 224 176 56 port 3000 timestamp 1460956559092506 value cr endpoint servicename ui app ipv4 10 225 145 5 port 3000 timestamp 1460956559093673 value ss endpoint servicename servicea ipv4 10 224 176 56 port 3000 binaryannotation key http status value 200 endpoint servicename ui app ipv4 10 225 145 5 port 3000 key http uri value v1 study environment assignment endpoint servicename ui app ipv4 10 225 145 5 port 3000 key sa value 1 endpoint servicename servicea ipv4 107 23 29 208 port 443 debug false traceid bd8283279c35f5e1 name get i d da98a54de4632943 parentid bd8283279c35f5e1 timestamp 1460956559126139 duration 18402 annotation timestamp 1460956559126150 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956559144538 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d 7e9f40c9e40fb32c parentid bd8283279c35f5e1 timestamp 1460956559158500 duration 20238 annotation timestamp 1460956559158507 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956559178735 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d ca5621cc4f030223 parentid bd8283279c35f5e1 timestamp 1460956559196350 duration 18378 annotation timestamp 1460956559196358 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956559214726 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d 546c80de2ee3b593 parentid bd8283279c35f5e1 timestamp 1460956559228957 duration 20401 annotation timestamp 1460956559228964 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956559249353 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d 7d18c775164aea6e parentid bd8283279c35f5e1 timestamp 1460956559272001 duration 33826 annotation timestamp 1460956559272011 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956559305823 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false traceid bd8283279c35f5e1 name get i d cc96b698a733d884 parentid bd8283279c35f5e1 timestamp 1460956559319752 duration 20110 annotation timestamp 1460956559319759 value sr endpoint servicename servicea ipv4 10 224 253 240 port 3000 timestamp 1460956559339859 value ss endpoint servicename servicea ipv4 10 224 253 240 port 3000 binaryannotation debug false |
openzipkinzipkin | how to get the trace from brave client filter for inter service call | Question | hi currently I be do a poc on zipkin for java web service implement in jersey2 I want to trace the call make by one web service application to other application the call be make by http use jersey2 client and also by httpurlconnection I have follow follow step 1 2 web service application jersey2 both web xml update with com github kristofa brave brave jersey2 for jersey config server provider package 2 web service a be call a rest api of web service use jersey client at one place and at another by httpurlconnection 3 create a jersey client to execute the web service a api which internally call we d service b api before execute the rest api register the brave request and response filter 4 start zipkin collector service zipkin collector service use dev configuration collector dev scala 5 start zipkin query service zipkin query service 6 start the zipkin web ui service 7 now execute the jersely cleint create at step 3 once with loggingspancollector and once with zipkinspancollector I want to confirm if the above step be sufficient to tarce the inter service call http call or I need to do some additional step in the trace log I could not see the trace for web service b trace log when there be no call to web service b from a span trace i d 3225959038768661911 name post i d 3225959038768661911 annotation annotation timestamp 1459937704152000 value cs host endpoint ipv4 172273347 port 0 service name pet store annotation timestamp 1459937705378000 value cr host endpoint ipv4 172273347 port 0 service name pet store binary annotation binaryannotation key http uri value 68 74 74 70 3a 2f 2f 6c 6f 63 61 6c 68 6f 73 74 3a 38 30 30 32 2f 61 70 69 2f 70 65 74 annotation type string host endpoint ipv4 172273347 port 0 service name pet store trace log when there be a call to web service b from a span trace i d 1045224323842629647 name post i d 1045224323842629647 annotation annotation timestamp 1460021984181000 value cs host endpoint ipv4 172182617 port 0 service name pet store annotation timestamp 1460021988212000 value cr host endpoint ipv4 172182617 port 0 service name pet store binary annotation binaryannotation key http uri value 68 74 74 70 3a 2f 2f 6c 6f 63 61 6c 68 6f 73 74 3a 38 30 30 32 2f 61 70 69 2f 70 65 74 annotation type string host endpoint ipv4 172182617 port 0 service name pet store binaryannotation key http responsecode value 35 30 30 annotation type string host endpoint ipv4 172182617 port 0 service name pet store thank anuj |
openzipkinzipkin | ui work splash stick on span view screen | Bug | I see this when run ui from the query service I e 1 37 screen shoot 2016 04 04 at 11 24 30 am the console error trace look like uncaught typeerror can not read property openchildren of undefined anonymous function trace js 94 jquery extend each jquery js 365 anonymous function trace js 92 jquery extend each jquery js 365 expandspan trace js 80 filteradde trace js 74 anonymous function trace js 273 |
openzipkinzipkin | high level step to instrument a java application and get result on zipkin ui | Question | hi I have recently setup zipkin and zookeper as per the documentation I have also brave library for java however I be struggle to find anywhere what will be the next step be there any documentation or reference that I can follow to instrument a sample application if someone who have already instrument for java jersey2 can provide the high level step sample reference would be of great help sorry if this be a duplicate I couldn t find any related open request |
openzipkinzipkin | don t distribute non shadow jar of collector service and query service | Bug | when people download the wrong jar the ui asset win t be bundle neither will any of the jar dependency I think mansu have this problem get the wrong jar or so it seem from the gitter conversation if there s no good reason why we re publish the not shadow jar maybe we could just as well not distribute they in the first place abesto be it possible to configure gradle so it doesn t build the jar only the shadow jar or at least doesn t upload the jar |
openzipkinzipkin | first index page be hit api v1 span servicename undefined | Bug | when first load the zipkin ui test with zipkin query 1 39 docker container it seem that the javascript be try to load span with undefined servicename |
openzipkinzipkin | proposal drop scala backend implementation in favor of zipkin java | Enhancement | the zipkin java project be create out of issue 451 463 it be now approach version 1 0 and result in question about the future of the scala implementation this issue will enumerate opportunity concern and any action to take on this topic pro zipkin java have more feature it have a no dependency library for example in instrumentation project custom collector or spark fine grain logging of dependency linking and collection stage fine grain and test health check system collector metric break out by transport integration test that invoke the package exec jar elasticsearch storage maintainer can do more if only support one server framework currently there s a lot of distract work port code and test and release both project docker and documentation be confusing when there be more than one authoritative process by cut scala zipkin s few maintainer we can progress more with the same time and cause less confusion the java build be more stable due to less dependency the scala project use more dependency include the unreliable maven twttr com repository which often break the build with http 503 error java implementation be make from well understand technology the server be base on spring boot which be a popular framework and well understand from a configuration pov the scala process be host on a relatively rare framework and literally load scala file to configure the process java implementation can be deploy as a single process while not require you can enable all feature include kafka in the same process java implementation have more people participate in difficult topic topic include modularity dependency management and configuration have have fast progress in the java project elasticsearch be write in 1 week and work both in the stock server and one base on armeria a netty rpc service this be in contrast to year of request in scala with no action con no precanned profile like collector some may desire a profile like collector include the ability to receive but not query span scala implementation battle harden at twitter performance availability implication a lot of zipkin code isn t use at twitter ex cassandra kafka but finagle scribe be those already use zipkin collector may have high confidence in finagle scribe vs fb swift use in zipkin java |
openzipkinzipkin | favicon ico miss from zipkin ui | Bug | seem like the zipkin ui be miss a favicon ico one we can use can be take from |
openzipkinzipkin | allow specification of start time in web search ui | Enhancement | sometimes it s nice to be able to narrow down your search a bit more the absence of this feature always seem a bit odd |
openzipkinzipkin | feature request please add an ability query with startts and not just endts | Enhancement | so the zipkin ui have a filter for the end timestamp be it possible to add a way to also query by the starting timestamp one example be if I be look for new trace I don t want to see trace that be 12 hour old so I would be able to provide a stop point in history not to pull those old trace |
openzipkinzipkin | zipkin ui consider open a new tab when click on a trace | null | when navigate from the find trace view to a trace you be fine navigate back be a problem as the result might be different gena01 suggest open a new tab on trace detail so that the find view be stable when the trace click isn t the one you want |
openzipkinzipkin | zipkin ui add hash to filename of app min js css map | Enhancement | the inside of the jar list look like this after the change meta inf meta inf manifest mf zipkin ui zipkin ui 448c34a56d699c29117adc64c43affeb woff2 zipkin ui 614fad616d014daf5367e068505cad35 png zipkin ui 89889688147bd7575d6327160d64e760 svg zipkin ui 8b55a822e72b8fd5e2ee069236f2d797 png zipkin ui app 8f28873f153b2b06e7ab min css zipkin ui app 8f28873f153b2b06e7ab min css map zipkin ui app 8f28873f153b2b06e7ab min js zipkin ui app 8f28873f153b2b06e7ab min js map zipkin ui e18bbf611f2a2e43afc071aa2f4e1512 ttf zipkin ui f4769f9bdb7466be65088239c12046d1 eot zipkin ui fa2772327f55d8198301fdb8bcfc8158 woff zipkin ui index html |
openzipkinzipkin | feature request update response header in zipkin query for zipkin ui asset | Enhancement | please add update response header to include cache control and or expire header when serve zipkin ui static file I see they be reload on every page click also seem like 8b55a822e72b8fd5e2ee069236f2d797 png which might be a favicon return the follow response header text content type image png charset utf 8 this be an image so no need for charset |
openzipkinzipkin | zipkin web 1 37 and zipkin query 1 38 web ui return error in web console | Bug | I be unable to use the late 1 37 docker container for zipkin web or just try zipkin query 1 38 with the web ui when I select a service from the dropdown list I be see nothing load and error in web console in both firefox and chrome firefox report too much recursion chrome report text lodash js 13824uncaught rangeerror maximum call stack size exceed object anonymous function lodash js 13824 anonymous function tracesummary js 15 6totalservicetime tracesummary js 161 anonymous function tracesummary js 161 totalservicetime tracesummary js 161 anonymous function tracesummary js 161 totalservicetime tracesummary js 161 anonymous function tracesummary js 161 totalservicetime tracesummary js 161 anonymous function tracesummary js 161 |
openzipkinzipkin | document improve zipkin ui route rule | Enhancement | the readme be available at before merge x merge 1049 this pr be base off of that x add reasoning on the packaging format choice to readme maybe eirslett could explain it well x verify zipkin query serve index html I can t compile it locally atm host method initially this pr describe four way of host zipkin ui asset 1 programatically read index route json and set up explicit route base on it example zipkin web l92 l95 pro serve index html for exactly the right route con need active server component suggest method when embed in a server application 2 match all unknown path and instead of return 404 route they back to index html example zipkin query pro never need to touch the routing rule again con no 404 may interfere with other route definition suggest method for static webserver 3 a variation on the above only serve index html for path where the last separate segment doesn t have a in it pro never need to touch the routing rule again have 404 for path with an extension con the heuristic be not very robust 4 manually hard code all the route that need to serve index html example zipkin java pro doesn t interfere with other route definition correct 404 con need update when new ui endpoint be add to zipkin ui after discussion see in line comment the conclusion be that we re drop 1 because let s embrace that we re a single page app drop 2 because it s an inferior version of 3 option 3 be directly support for example by finatra drop 4 because it s a temporary hack in zipkin java that ll go away let s not propagate it |
openzipkinzipkin | ui make expand all the default view for a trace | Question | transfer one of our internal ticket many user have ask for trace to be fully expand by default unless it exceed a certain n of span especially new user often get confuse by see incomplete trace by default |
openzipkinzipkin | deprecate zipkin ui web reorganize webpack structure | Enhancement | close in favor of 1049 same changeset initially but on a share branch |
openzipkinzipkin | load configuration from config json | Enhancement | rename the config js endpoint return json instead of set window config this be more flexible but also somewhat more involved since we have to load configuration asynchronously from the client code also add source map generation for js and css easy debugging |
openzipkinzipkin | feature more es6ify | Enhancement | a pretty big diff this be a bunch of cosmetic change to the js code it s all es2015 now I ve apply airbnb s code style guide with a few tweak we run linte with npm run lint abesto do you think it should be part of the gradle build set up with some kind of dependency |
openzipkinzipkin | span should get child ids | Bug | it look like there be a bug in rendertrace probably an oversight after a refactoring in the web ui the datum modelview prepare for trace render every span have a list of child child span id this list only contain a single element the i d of the span itself I m sure the intention be to get the actual list of child it look like this may have fix 999 at least in some case |
openzipkinzipkin | thrift less cassandra schema | Enhancement | both michaelsembwever and yurishkuro hint that a thrift less cassandra schema might be useful let s evaluate what such a schema might look like and a transition plan |
openzipkinzipkin | master branch duration in waterfall show incorrectly | Question | I run zipkin from source commit be 95c017e4eaf372060fe2e587ea6638c2a8448775 see below screenshot image duration be calculate by server receive server send instead of client receive client send |
openzipkinzipkin | render the index default page completely client side | Enhancement | the default page use to leverage a modelview endpoint on the scala server to generate a model to pass into its mustache template the logic from that modelview endpoint be now port to javascript so that all the pre render business logic happen on the client side since this be a high risk change the modelview endpoint still exist on the server every time we request a page the ui will actually fetch a process version from that endpoint as well and compare the scala version with the javascript version the idea be that we can compare output of the two and if there s a difference we can assume that s a bug in the javascript port the endpoint and the ui logic to compare implementation will soon be remove after test please don t merge this pr yet I really need people to test this change manually it would be great if you can checkout this branch start zipkin web and look at a few trace summary not trace but the list of trace summary search result on the front page in the browser debugger there s a console log thing that log whether the javascript implementation give the same result as the scala implementation give the same input I would love to hear if anybody experience a situation where the two implementation produce different output and in that case what the input be if you ve try run zipkin web with this change and you find no bug be sure to 1 the pr whenever we re sure that most bug be clear out I ll update with another commit that remove the automatic scala vs javascript comparison on every page load |
openzipkinzipkin | 1 33 2 collapse button in trace detail page not work | Duplicate | use the datum in tracegen when click the button at the left of service natoque the service do not expand before click after click |
openzipkinzipkin | the total time of each span seem incorrect | Question | in the summary page home page zipkin will show a summary time of each span below each trace we have many redi span each cost 300u but in the summary page it show redis 232 5ms 5ms 232 21 5us so I think may be the summary data be incorrect |
openzipkinzipkin | add health endpoint to zipkin query and zipkin web | Enhancement | various platform have admin instance that expose health and twitter s do so on a different port by introduce a health endpoint we can start a portable conversation about health that doesn t rely on platform convention or open more port on the firewall the use case for this be load balancer health check and the zipkin javascript ui see 994 |
openzipkinzipkin | support heath endpoint | Enhancement | in our testing tool we be check that the zipkin query server be run but do a head request to the api this will return 404 but if we do a get request it will respond normally with 200 that s an extra burden to be send datum when it do need to and the api should be respond to head also optionally or in addition the zipkin query server could have some status or similar endpoint to check its health |
openzipkinzipkin | fix ui navbar highlighting | Bug | the find a trace menu element be hard code to always be select this fix create a navbarui component which listen to navigate event and highlight the correct menu item depend on the route |
openzipkinzipkin | fix ui bug trace placeholder wasn t show | Bug | after the ui refactore a bug in the handling of queryresult surface this mean the placeholder that be show before you have make any query be never render this fix add a querywasperformed property that be compute client side the structure of the json return from the server be also simplify a bit yurishkuro I think this should fix the bug you report |
openzipkinzipkin | improve ux when multiple binary annoation exist in the same span with the same key | Bug | why do the same binary annotation pop up twice in zipkin it be possible to have multiple host in the same span each of they can log the same binary annotation tag ex http status when they match this look like a duplicate when they don t match this could be helpful ex the client perceive an error even if the server didn t perceive one single host span by definition don t have this problem since only one host ever log anything in a span here s an option yurishkuro discuss about uncover why it look like a dupe basically if we add a column for the origin of the value it might make more sense I ve paste some other thought of his below as well image another common scenario I ve see with same key be the http url which be send by both client and server and the value may or may not be the same if we adopt the suggestion that spanid spankind emittingservice combination be unique which I think imply ditch core annotation and move the endpoint on the span itself then seemingly identical binary annotation may still be distinguishable by the service that emit they |
openzipkinzipkin | improve cache heard of the ui s asset | Question | zipkin ui render very fast once all the asset be retrieve app min js have max age 600 must revalidate cache header and it s be send every time that s the main source of wait on the ui for I at least I think add well cache header will make wonder for ux eirslett probably interested |
openzipkinzipkin | remove all server side mustache code from zipkin web | Enhancement | the server now only deliver a static index html page which in turn load javascript css to run the application remove mustache dependency from gradle add index html container page for host the application add new flight ui component common for render the application layout the environmentui component be now render by the common component use event to trigger loading of servicename spanname to populate dropdown |
openzipkinzipkin | ui refactore | Enhancement | my plan be to refactor zipkin web to that the web gui for zipkin become a pure html javascript css frontend on top of zipkin query s rest api a lot of work have already be do introduce webpack as a build system for the frontend move bit and piece of logic from server to client some more to be do render all the mustache templates client side but let some business logic happen on the server the reason there will temporarily be some stuff be do server side be that there s a lot of complexity and it would be very risky to rewrite it all in one go most of that remain server side stuff will be json from query server transform into other json that can be consume by mustache template webpack have also introduce babel a javascript cross compiler that will let we use new es2015 syntax feature and it compile it back to es5 javascript that all browser support we should migrate the code base gradually towards es2015 in the place it make sense the end result of the refactoring would be to move all remain business logic from zipkin web into the browser bundle rewrite it from scala to javascript in the process then we can get rid of zipkin web entirely and serve the static asset directly from zipkin query or apache httpd or nginx |
openzipkinzipkin | move all mustache render to the browser | Enhancement | this mean the zipkin web backend be now only basically a json api deliver static asset there s a lot of complexity go on in the transformation from trace span json datum to datum that be feed into mustache template these transformation be still write in scala but host at modelview endpoint ideally they should be port to scala some code have to be move around event handler must be bind after the html be render which now happen at a later stage than it use to when it be server side render |
openzipkinzipkin | ui the dependency be not present at all for service | Question | for the follow json receive from the query server traceid 5a646801ed9b6dc5 name http reservation name i d 5a646801ed9b6dc5 timestamp 1 455352726564e 15 duration 866000 annotation endpoint servicename reservation service ipv4 192 168 1 104 port 7074 timestamp 1 455352727203e 15 value sr endpoint servicename reservation service ipv4 192 168 1 104 port 7074 timestamp 1 455352727386e 15 value ss binaryannotation key http host value localhost endpoint servicename reservation client ipv4 192 168 1 104 port 7073 key http method value get endpoint servicename reservation client ipv4 192 168 1 104 port 7073 key http path value reservation name endpoint servicename reservation client ipv4 192 168 1 104 port 7073 key http url value http localhost 7073 reservation name endpoint servicename reservation client ipv4 192 168 1 104 port 7073 traceid 5a646801ed9b6dc5 name http reservation i d e1592db4f9667be9 parentid 5a646801ed9b6dc5 timestamp 1 455352727199e 15 duration 191000 annotation binaryannotation key http host value 192 168 1 104 endpoint servicename reservation service ipv4 192 168 1 104 port 7074 key http method value get endpoint servicename reservation service ipv4 192 168 1 104 port 7074 key http path value reservation endpoint servicename reservation service ipv4 192 168 1 104 port 7074 key http url value http 192 168 1 104 7074 reservation endpoint servicename reservation service ipv4 192 168 1 104 port 7074 the dependency for those two application be not present at all also servicename be miss for one application miss service name |
openzipkinzipkin | ui span be shift to the left their relative time be negative | Question | image span bug json span bug txt |
openzipkinzipkin | bugfix make sure that the span be wrap in jquery | Bug | I m not sure if it s affect the end user experience in any way but we get a javascript exception that block execution when you click the expand all button to show all span in a trace you must open the browser debugger to see it apparently javascript s weak type system have be abuse leveraged self spansbyservice be a map from service name to list of span ids string but that list of ids will eventually be convert lazy initialise into a list of jquery object that wrap the span dom element the lazy initialisation mostly happen inside getspansbyservice I think we should refactor the code so that it create jquery object up front for now this patch should fix the exception |
openzipkinzipkin | how should we handle single host rpc span in the ui | Enhancement | at least htrace opentracing google cloud trace and sleuth use single host span for example client send and server receive be in different span base on the client or the server tracer zipkin s v1 model be design to have client and server in the same span hence the need for the annotation cs sr ss sr zipkin v2 will probably move to single host per 939 zipkin s ui be design for when a single span contain both client and server side ex there s a single bar for a client server span and there be dot for each of the major client or server event such as client receive or server send this cut the visual depth of a trace quite a bit conversely single host span have a separate span for a client or a server this mean that a server span be visually below its client counterpart for the same amount of rpc this be double the bar vs client server span do we need any ui work wrt single tracer rpc span or be status quo ok |
openzipkinzipkin | ui service name not present for some span | Question | hello in our e2e test project we re test integration with zipkin for some of the create span the ui be not present the service name zipkin I ve do an analysis of the json from zipkin and it seem to be valid most likely there be a bug in the ui how to reproduce either see the attach json zipkin json txt or execute this for linux git clone cd brewery bash runacceptancetest sh t sleuth k and then check localhost 8080 for zipkin ui |
openzipkinzipkin | ui frozen highlighting in the navigation bar | Bug | maybe someone know how to quickly fix it in mustache l23 html find a trace dependency the class active be always hardcode on the find a trace menu item even if the page be show dependency |
openzipkinzipkin | zipkin v2 span model | Enhancement | we should not try to do all thing as zipkin v2 the large amount of interesting thing may lead we to never even start I believe that if we pare down do basic create a model that only include feature that work in zipkin 1 backend we can start and finish something for example zipkin 1 only really support string string binary annotation let s make they string string for v2 even if they become fancy later zipkin v1 allow contributory span where instrumentation be allow to place multiple host into the same span let s prevent that and use a portability technique google cloud trace do google cloud trace allow multiple span with the same i d but they vary on span kind for example core server annotation really imply a span kind server where sr be span startts and ss be span endts we could probably literally use the google cloud trace model as zipkin v2 as it be almost exactly what we need tracespan if we do something like this I m confident we can get v2 out in relatively short order and still not prevent more interesting and severe change from occur in later version of zipkin |
openzipkinzipkin | rfc fix anormdb dependency | Bug | as describe in 929 shortly anormdependencystore do not report dependency that have intermediate span between rpc like span1 sr ss span2 call span3 cs sr ss cr dependency 1 this change also add dependency use client addr and server addr like span1 can sr ss dependency 1 span2 call span3 cs cr sa dependency 2 917 case however it doesn t pass test yet by 2 reason one be the same to 917 dependency headlesstrace the other be sort of result it report cansearchforintervalsbesidestoday fail org scalatest exception testfailedexception list zipkin query zipkin jdbc 1 zipkin web zipkin query 1 be not equal to list zipkin web zipkin query 1 zipkin query zipkin jdbc 1 dependencystorespec scala 248 I wonder it must be topologically sort so I want to fix test to ignore order of result instead of implement topological sort in the store also this change include complex costly sql I want comment too |
openzipkinzipkin | rfc extend dependencylink datum model to include infer instrumentation | Enhancement | background adoption be one of the big challenge to successful deployment of distribute tracing in an organization developer often need to upgrade dependency of their service and change configuration in the good case or even change the application code in the bad case therefore automatic tracking of adoption stat be very important for distribute tracing team the dependency graph partially provide these stat however the edge in the graph sometimes represent a link between two service where only one of the service be fully instrument for trace and the other service name be simply infer from rpc field cf 914 this be useful information to track in the dependency datum model since that information be explicitly available in the trace that be aggregate into dependency graph proposal add an optional instrumentation field to dependencylink thrift thrift struct dependencylink parent service name caller 1 string parent child service name callee 2 string child 3 moment obsolete duration moment call make during the duration of this link 4 i64 callcount histogram which service s actually submit span datum that result in this link 5 optional instrumentation instrumentation describe which service s submit span datum that result in a dependencylink enum instrumentation both parent and child service submit span datum parent and child only the parent service submit span datum the child service name be infer from it parent only only the child service submit span datum the parent service name be infer from it child only scala model in the object model we need a way to merge two link with the same parent child pair e g when merge dependency set from several day the proposal be to give parentandchild a priority over the other two value scala seal trait instrumentation def other instrumentation instrumentation case object parentandchild extend instrumentation override def other instrumentation instrumentation this case object parentonly extend instrumentation override def other instrumentation instrumentation other match case parentandchild other case childonly parentandchild case this case object childonly extend instrumentation override def other instrumentation instrumentation other match case parentandchild other case parentonly parentandchild case this |
openzipkinzipkin | how to track async span | Enhancement | so I think we ll need to think about async or background task a little bit more at this moment if you have let say service a calling service b and service b use a background task to send an email you can do two thing the background task be in the same trace the background task be in a different trace the second option be good ui wise you can search for all those trace but you lose the connection to individual request the first option have the problem that if the background task be slow it look like even servicea be slow we have try and even select by name or with filter in the ui trace will always show fully and the total time be the time of the slow ideally the solution would be to mark that those span happen in the background be in fact background or async span and that we do not want to count they for the calculation of how much the response take the same solution would work for ajax request which probably you do not want to compute they as the response time but still you want to have they connect to the request that originate they technically maybe be a binary annotation add to the span to mark it as async and the server would need a bunch of change do you think that be a good idea at all or it should be solve in some other different way dankosaur dsyer |
openzipkinzipkin | track indirect service dependency | Enhancement | imagine a case where some route layer like haproxy be add its own span to pass through trace this make the dependency graph not very interesting as instead of a b link for any pair of a and b we would get a haproxy and haproxy b such edge have merit but they completely lose the a b dependency even though the datum in the span have it we can extend the dependency datum structure to still allow link a b to be present but mark as indirect the ui can be extend to toggle between show hide indirect link or maybe there s even well way open to suggestion for example when you mouse over a service in the diagram it dim unconnected service and highlight directly connect one for service a above it would highlighght haproxy but dim b it would be nice in that mode to also show 2 hop service e g highlight b as well but in different color |
openzipkinzipkin | fix select end time no effect | Bug | end time handling seem to be corrupt and this change fix 786 |
openzipkinzipkin | why do span servicename give priority to serveraddress | Question | I agree that server side should take precedence over client side but it seem the coreserver annotation should be the top priority not serveraddress binary annotation in the current form if 10 client call the same service but all decide to call it differently in serveraddress it will appear under 10 different name and if the service owner want to find trace in their service they need to search under 10 name but if the coreserver annotation be the first priority it will always have the same name |
openzipkinzipkin | improve ui control behaviour before js have load | Bug | I find often that before the js have fully load execute on the query result page I ll mistakenly click on a ui control such as the sort order dropdown and find that it do nothing imo it would feel much nice to do one of the follow make it clear to the user that any interaction with the page before js have load will not work disable the ui control handle any change to the ui at the point when the js be load probably more complex but would arguably be the good solution I suppose this could be implement by check if any ui control value be different from those provide in the query param and perform the appropriate action if any difference be find |
openzipkinzipkin | single sided rpc span be not recognize for dependency diagram | Enhancement | specifically in cassandra implementation in zipkin dependency spark not sure about other store there be situation where only the server or the client be instrument yet it be able to identify the peer and emit a clientaddress or serveraddress annotation within the single sided span I want to add a test to the dependecystorespec that will expect the store to return a parent child link even for this span I also want to extend the dependencylink model to include a boolean flag singleside that will indicate that even though this link have be detect from the span the instrumentation be not actually complete which be useful for adoption track stat such link can be display in the ui as a dash line at minimum I would like to have this flag in the thrift class fyi the change be mean to pass these two test scala in some case an rpc call be make where one of the two service be not instrument however if the other service be able to emit sa or can annotation with a service name the link can still be construct test def getdependenciesonlyserverinstrumente val client endpoint 127 24 1 9410 not instrument client val server endpoint 127 24 2 9410 instrument server val trace applytimestampandduration list span 10l get 10l annotation list annotation today 100 1000 constant serverrecv some server annotation today 350 1000 constant serversend some server binaryannotation list binaryannotation constant clientaddr true some client processdependencie trace val traceduration trace duration trace get result store getdependencie today 1000 sortby parent should be list new dependencylink not instrument client instrument server 1 in some case an rpc call be make where one of the two service be not instrument however if the other service be able to emit sa or can annotation with a service name the link can still be construct test def getdependenciesonlyclientinstrumente val client endpoint 127 24 1 9410 instrument client val server endpoint 127 24 2 9410 not instrument server val trace applytimestampandduration list span 10l get 10l annotation list annotation today 100 1000 constant clientsend some client annotation today 350 1000 constant clientrecv some client binaryannotation list binaryannotation constant serveraddr true some server processdependencie trace val traceduration trace duration trace get result store getdependencie today 1000 sortby parent should be list new dependencylink instrument client not instrument server 1 as I be not familiar with other span store implementation the new test will be disable there for now any objection suggestion |
openzipkinzipkin | do not replace empty string with unknown and do not return they to ui | Bug | fix 896 |
openzipkinzipkin | ui should not prefer empty service name in annotation over non empty one | Bug | when instrument outgoing http call the target service name be not always know so the client can emit serveraddress annotation with the correct host port but empty service name this result in the span display in the ui with service name of unknown even though when the server be instrument it emit a valid service name in sr sr annotation propose fix do not replace empty service name with unknown string as it make further processing filtering difficult make sure the empty service name do not take precedence if the span contain annotation with valid service name |
openzipkinzipkin | aggregate tab not visible and how to show dependency graph | Question | hi I be unable to see aggregate tab in zipkin ui be it some sort of configuration or it be remove |
openzipkinzipkin | limit select statement in getspanname to avoid timeout | Enhancement | some user can abuse zipkin s design and insert crazy number of span name underneath each service name this will initially manifest itself as a problem through timeout when try to read all span name for a give service name back out again the only purpose to the span name table be for the ui component display all span name no one be go to scroll through such a large list limit the select statement as this patch do be the small change in the code to tackle the immediate problem it may be that the frequent writing to the same service name row create hot partition and cause problem to the cassandra cluster this problem can be tackle when and if it arise ref |
openzipkinzipkin | dependency view should deduplicate edge | Bug | due to the fact that the ui do not specify the startts param call the query service duplicate edge might be add to the graph which result in an exception on the js side uncaught error graph already have edge servicea serviceb the problem be cause by run the dependency job on different day in term of the ui assume a range of more than one day be display the edge count could be add and one unique edge with the sum would exist in the graph it could also be implement on the java side probably in cassandradependencystore to be analogous to anormdependencystore which do the sum operation and provide unique edge |
openzipkinzipkin | discrepancy in the number of span per trace report in the ui | Bug | search result show that a trace have 4 span servicea 1x serviceb 3x however when go into detail of that trace there be only 3 span there be indeed 4 record in the database because one of the span be rpc and thus have two side submit to storage separately the front page may not be do the merge of the span properly |
openzipkinzipkin | add cache control to public css and js | Enhancement | I m not scala programmer but look at this I think it should work the ui be slow look at the cause it be because we have lot of css and js and they be send everytime adriancole please run test xd |
openzipkinzipkin | configurable cassandra driver s read timeout | Enhancement | on large dataset I ve notice frequent com datastax driver core operationtimedoutexception default timeout be 12 second which be quite small please add configuration flag for this parameter for query service so it can be overriden or provide a way to override it in config |
openzipkinzipkin | can not see late trace when many in the db | Bug | how the ui work now seem to work stage 1 search here you specify the service name endtime and limit stage 2 ordering of result once there be result there be an ordering dropdown to order these result I have 30 000 trace for a give service these trace have be send the last 1 month or so I want to see the late trace the one send the last couple of day or so but it be impossible if I try to get more than 1 000 trace in the ui the connection will timeout if I get 1000 trace then those seem to be the old so I can not see anything that have happen recently look at the ui in fact I can not even know if zipkin keep receive trace or no and also I do not want to see 1000 trace in the ui possible solution add a start date to the ui or the ordering dropdown be there before click find and the ordering happen before the call to the db or both |
openzipkinzipkin | introduce local component lc binary annotation and example use case | Enhancement | local span have be discuss at length but have never be formalize this formalize how to tag a span as local include rationale documentation to help demonstrate the point zipkin web and zipkin query use this for bootstrap trace fix 808 here be example local span from finatra and twitter server bootstrap |
openzipkinzipkin | move clock skew correction to spanstore for portable testing | Enhancement | clock skew correction be vital for the ui to appear properly particularly in the current model this be because there s no concept of sequence of span except by timestamp the type of clock skew adjustment zipkin do be shift child span to a later timestamp when they appear to happen before their parent formerly there be no test to show span store actually do this this mean you can create implementation which appear portable but be not useful as they result in a poor ui this move the implementation of clock skew correction to the common package and introduce a new test in spanstorespec which assert what expectation be without limit how implementation correct clock skew for example this test will pass regardless of whether clock skew be correct at storage or query time |
openzipkinzipkin | support local span via a local component lc binary annotation | Enhancement | currently it be tribal knowledge that local in process span exist we need a way to designate span as local and offer insight into how to use they this can lead we to answer what status quo could be on this and how they might be test firstly local span need to be annotate with a namespace that can be use like service a discussion on github suggest use a local component lc binary annotation for this here s a json view json key lc value encryption service endpoint servicename proxy ipv4 1 2 3 4 port 345 you ll notice that the binary annotation hold almost all information need to identify the context of the span it contain a searchable annotation which contain its namespace in this case encryption service via the exist contract of binary annotation it also include the service that contain it in this case proxy what s leave be the skeleton of the span how do we capture the duration of the span I m aware of three option re use cs cr annotation make up new annotation for the bookend of a local span piggy back off span startts duration 807 tl dr be that I think the latter be good if we re use cs cr annotation we avoid logic around dependency graph where server annotation be special case in other word we avoid accidentally clutter it with local span we can also hang off an exist explanation where we use cs cr to address uninstrumented system it might be the least code change json traceid 0000000000000005 name encrypt i d 000000000000029a parentid 0000000000000002 annotation timestamp 60 value cs endpoint servicename proxy ipv4 1 2 3 4 port 345 timestamp 100 value cr endpoint servicename proxy ipv4 1 2 3 4 port 345 binaryannotation key lc value encryption service endpoint servicename proxy ipv4 1 2 3 4 port 345 if we make up new annotation for the bookend we avoid add more cognitive weight to cs cr which be already confusing instead all two character annotation key prefix l explain local span and those not interested simply don t read their doc we do have to add more constant and affect the instrumentation more while this be nice for documentation containment it could look overwhelming json traceid 0000000000000005 name encrypt i d 000000000000029a parentid 0000000000000002 annotation timestamp 60 value ls endpoint servicename proxy ipv4 1 2 3 4 port 345 timestamp 100 value lr endpoint servicename proxy ipv4 1 2 3 4 port 345 binaryannotation key lc value encryption service endpoint servicename proxy ipv4 1 2 3 4 port 345 if we piggy back off span startts duration 807 local span really only need span metadata and the single binary annotation this would result in less storage and cognitive burden as we no long need to think about annotation in order to understand the duration of a span this be the most dramatic departure from exist core annotation so need to be specifically support in ui and presentation json traceid 0000000000000005 name encrypt i d 000000000000029a parentid 0000000000000002 startt 60 duration 40 binaryannotation key lc value encryption service endpoint servicename proxy ipv4 1 2 3 4 port 345 |
openzipkinzipkin | enable duration query by promote span duration timestamp | Enhancement | a commonly request feature be query by duration for example show trace long than 1 minute while the ui support ordering by duration it be a game of chance whether the trace return happen to be long or not even if you put a limit of 1000 you still may not find the long trace as your storage system may have a million trace in it the path propose be top level span timestamp and span duration which formerly exist in the scala and mustache model but not in the thrift we could implement duration query without change thrift for example this could be an implementation detail of each storage system however this limit testing to side effect and keep essential detail in scala code make portability hard also we have other reason to top level these field local span can be small 808 non rpc span can t be support without artificially log annotation like cs cr reduce redundant span duration logic in several scala class and mustache template dependency graph link can cheaply add duration aggregate here s a convenience paste of the definition of span timestamp and span duration java microsecond from epoch of the creation of this span this value should be set directly by instrumentation use the most precise value possible for example gettimeofday or sync nanotime against a tick of currenttimemillis for compatibilty with instrumentation that precede this field collector or span store can derive this via annotation timestamp for example server recv timestamp or client send timestamp this field be optional for compatibility with old datum first party span store be expect to support this at time of introduction 10 optional i64 timestamp measurement of duration in microsecond use to support query this value should be set directly where possible do so encourage precise measurement decouple from problem of clock such as skew or ntp update cause time to move backwards for compatibilty with instrumentation that precede this field collector or span store can derive this by subtract annotation timestamp for example server send timestamp server recv timestamp if this field be persist as unset zipkin will continue to work except duration query support will be implementation specific similarly set this field non atomically be implementation specific this field be i64 vs i32 to support span long than 35 minute 11 optional i64 duration |
openzipkinzipkin | service dropdown double | Bug | run v 1 19 2 on load home page the server log 14 01 27 215 cluster1 nio worker 2 info c t zipkin example servicenamecache cache return service list mezzannine minimez 14 01 27 216 cluster1 nio worker 2 info c t f h filter accessloggingfilter 127 0 0 1 30 oct 2015 18 01 27 0000 get api v1 service http 1 1 200 24 3 14 01 27 218 cluster1 nio worker 2 info c t f h filter accessloggingfilter 127 0 0 1 30 oct 2015 18 01 27 0000 get api v1 span servicename mezzannine http 1 1 200 11 5 and the service dropdown look correct screen shoot 2015 10 30 at 1 55 51 pm then select one service in the dropdown server log still look fine 14 01 48 518 cluster1 nio worker 2 info c t zipkin example servicenamecache cache return service list mezzannine minimez 14 01 48 519 cluster1 nio worker 2 info c t f h filter accessloggingfilter 127 0 0 1 30 oct 2015 18 01 48 0000 get api v1 service http 1 1 200 24 4 14 01 48 521 cluster1 nio worker 2 info c t f h filter accessloggingfilter 127 0 0 1 30 oct 2015 18 01 48 0000 get api v1 span servicename minimez http 1 1 200 14 5 but the ui double the list of service screen shoot 2015 10 30 at 1 56 19 pm reload to home reset the dropdown to correct state and the cycle repeat |
openzipkinzipkin | allow viewing of old dependency | Enhancement | while the dependency api accept a stop point endts 1445470517980000 the ui doesn t expose this this mean people make demo with can datum will see nada and have no mean to in the mean time folk use spigo need to re run their simulation before post to zipkin if you see nothing maybe it be just old datum cc adrianco |
openzipkinzipkin | too many connection to the mysql db | Bug | after run the query and collector for a while I can see the mysql db maxe out its connection error 1040 08004 too many connection it seem our setting be 120 connection I can see it slowly build up and never release even when there be no activity in query and probably almost none in the collector |
openzipkinzipkin | annotation query search doesn t work if end timestamp give of future | Bug | whenever I play with zipkin to track query realtime I change the end timestamp to tomorrow so that I need not update the timestamp every minute this work fine except for annotation query search step to reproduce 1 find an annotation query which give non zero result 2 change end timestamp to tomorrow 3 search with the same annotation query it return zero result |
openzipkinzipkin | authn support | Enhancement | in our deployment of zipkin some internal customer be ask for authentication support such that we can restrict the access on a per service level the reason for this be mostly that binary annotation may contain security sensitive information what I would be interested in be secure the query service via some form of auth and then modify web to prompt and forward that information thought |
openzipkinzipkin | dependency graph be incorrect | Bug | I have a test setup where traceproducerone sleep 0 5 then call traceproducertwo which sleep 0 2 and call traceproducerthree which sleep 0 3 clicking on the dependency graph I would expect traceproducerone traceproducertwo traceproducerthree instead I have the following graph which be incorrect as one never call three directly when I click on traceproducerone I get this which be incorrect |
openzipkinzipkin | revamp the zipkin data model | Enhancement | revamp of the datum model will be summarize here after discussion in a google doc |
openzipkinzipkin | doesn t work with amazon rd | Question | we ve be try to get the mysql storage to work on amazon rds with little success use docker zipkin we manage to use the mysql storage and it work fine change the mysql host to an rds instance send dummy trace and try to see they in the web ui by submit the search form end up in error such as query 1 java lang runtimeexception unexpectednullablefound columnname zipkin annotation end ts some end ts query 1 at scala sys package error package scala 27 zipkin query jar na query 1 at anorm sql anorm sql as anorm scala 472 zipkin query jar na query 1 at anorm sql class as anorm scala 292 zipkin query jar na query 1 at anorm simplesql as anorm scala 190 zipkin query jar na query 1 at com twitter zipkin storage anormdb anormspanstore anonfun gettraceidsbyname 1 apply anormspanstore scala 204 zipkin query jar na query 1 at com twitter zipkin storage anormdb anormspanstore anonfun gettraceidsbyname 1 apply anormspanstore scala 180 zipkin query jar na query 1 at com twitter zipkin storage anormdb db withrecoverableretry db scala 120 zipkin query jar na query 1 at com twitter zipkin storage anormdb db anonfun innewthreadwithrecoverableretry 1 apply db scala 147 zipkin query jar na query 1 at com twitter util try apply try scala 13 zipkin query jar na query 1 at com twitter util executorservicefuturepool anon 2 run futurepool scala 115 zipkin query jar na query 1 at java util concurrent executors runnableadapter call executor java 511 na 1 8 0 60 query 1 at java util concurrent futuretask run futuretask java 266 na 1 8 0 60 query 1 at java util concurrent threadpoolexecutor runworker threadpoolexecutor java 1142 na 1 8 0 60 query 1 at java util concurrent threadpoolexecutor worker run threadpoolexecutor java 617 na 1 8 0 60 query 1 at java lang thread run thread java 745 na 1 8 0 60 in the console while the web ui itself be a white unstyled screen state error internal server error com twitter finagle nostacktrace unknown source we ve be try both mysql 5 6 and 5 5 no change the only difference between work doesn t work really be the db everything else be the same the schema etc be also the same adriancole any pointer would be highly appreciate this be hopefully the last blocker before we put zipkin to good use here cc jcarre mdsol |
openzipkinzipkin | find an alternative to dagre d3 for dependency view | Enhancement | accord to its homepage dagre d3 be no long maintain let s figure out what we should use instead cc eirslett |
openzipkinzipkin | spanname filter should consider character case in zipkin web ui | Bug | I m use finagle which might record it s rpc method as span name e g getresult in the zipkin ui there will be a low case getresult show in the dropdown when I select it I ll find nothing but if I manually change the url from like to I could get the right answer it look like we should not transform the span name into low case in ui for selection |
openzipkinzipkin | settle on startts endts and fix dependencie ui | Enhancement | troubleshoot subtle problem be unnecessarily difficult as we use many way to refer to the start and end of a span in epoch microsecond this settle on startts and endts which be the dominant name by do so it be cognitively easy as for example we can see that span endts relate directly to the query parameter endts as well as the upper bind of dependency endts this bump the version because it fix the dependency api endpoint to use normal query parameter as oppose to path expression for startts and endts this also rename the thrift field for dependency although that isn t break as name aren t encode in thrift serialization most importantly this fix the dependency ui which have be break for a while |
openzipkinzipkin | initial implementation of cache service name | Enhancement | ui service name dropdown be populate use ajax call to api service endpoint use servicename component http response to api service include cache control max age 300 must revalidate header which make browser return cache value of service name browser cache for api service call expire after 5 minute 300 second by default flag zipkin web servicenamecachetimeoutinsec can be use to set this cache timeout to a custom value |
openzipkinzipkin | fix span name selector | Bug | the span name selector ui be call gettrace instead of getspanname this happen during the switch to the json api this change fix that and clarify the confusing naming that be use in the web handler fix 748 |
openzipkinzipkin | span name selector ui break | Bug | with the merge of the json api branch the selector in zipkin web be break because it load datum from api span servicename xxx which return a nest array of span and then try to display it as if it be a array of string previously the datum be load from a span name endpoint which use getspanname in the span store to retrieve a list of span name this approach be superior because it allow the span store to execute an efficient query to just aggregate all span name across the trace the aggregate could even be store separately |
openzipkinzipkin | db error when try to display 1000 item in the ui 698 | Bug | quick fix to enable limit on query that use in and order by cassandra 6722 see |
openzipkinzipkin | clarifie contract of dependencystore and backfill test | Enhancement | this standardize on micro for dependency start end interval and backfill test in a way that doesn t flake on timezone it also clarify doc and converge the dependencysource sink as prior to our move to json in the near future we be go to migrate from thrift to json api this backfill test and revise the contract around dependency so that future work isn t confusing or difficult to implement portably dependency implementation be confuse use a combination of microsecond date millisecond time duration and calendar instance to do the job there be also incorrect doc confusing which of starttime or endtime be the default anchor of the interval this be very difficult to understand and easy to break as there be no test that cover for example what the result value should be it be also difficult to understand the difference between the dependency api and limitation of the cassandra implementation |
openzipkinzipkin | convert the communication between web and query to http | Enhancement | formerly zipkin s ui communicate with the query server via thrift rpc this mean that any implementation that support the ui need to run a thrift listener there s many problem with this thrift rpc isn t support in most framework hence a learning curve thrift rpc isn t support by loadbalancer so you have to go tcp the only thrift rpc trace implementation be custom to finagle finagle have their own tcp multiplexer which isn t support elsewhere bad than all of this be that the choice be a poor match to power a ui zipkin web literally convert these call to http json and there s no other consumer of the query api move this link to normal http json will make zipkin easy to maintain interop with and troubleshoot |
openzipkinzipkin | service dependency ui enhancement | Enhancement | each item might need to be split into separate issue but I want to just list they all in one place without some of these feature the dependency graph be practically unusable for large system e g 500 service 1 there should be a combobox with service name and select a service should act similar to what happen when you mouse over the service in the graph I e highlight connect service dim the rest 2 there should be a way button checkbox to trim the graph to only the service connect to a select service it can actually be a dropdown with distance I e trim the graph to service that be within specify number of hop from the select service currently highlight be only one hop 3 the edge be currently highlight on mouse over but not clickable click an edge should bring the same second level popup as when you click on a service first and then on one of its up downstream friend |
openzipkinzipkin | regex capability to ui annotation query search | Enhancement | if my http uri be foo bar bla asdf I should be able to search all trace which have value as foo be there already a way for it |
openzipkinzipkin | duplicate annotation show in ui | Bug | ui the timeline be weird because I be debug in the service location service but I get the same thing with a normal timeline when I don t do that date time relative time service annotation host 21 09 2015 06 00 30 10 000ms fleet location updater acquire 127 0 0 1 9007 21 09 2015 06 00 30 10 000ms fleet location updater acquire 127 0 0 1 9007 21 09 2015 06 00 30 11 000ms fleet location updater client send 127 0 0 1 9007 21 09 2015 06 00 30 11 000ms fleet location updater client send 127 0 0 1 9007 21 09 2015 06 00 30 18 000ms service location service server receive 127 0 0 1 9001 21 09 2015 06 00 30 18 000ms service location service server receive 127 0 0 1 9001 21 09 2015 06 00 35 5 018s fleet location updater release 127 0 0 1 9007 21 09 2015 06 00 35 5 018s fleet location updater release 127 0 0 1 9007 21 09 2015 06 01 30 59 968s service location service server send 127 0 0 1 9001 21 09 2015 06 01 30 59 968s service location service server send 127 0 0 1 9001 database mysql select from zipkin annotation where value sr and not service name like zipkin span i d trace i d span name service name value ipv4 port a timestamp 5601236271552241165 111825377246723855 message input service location service sr 2130706433 9001 1442829630633000 1 row in set 0 17 sec mysql select from zipkin annotation where service name fleet location updater and trace i d 111825377246723855 span i d trace i d span name service name value ipv4 port a timestamp 5601236271552241165 111825377246723855 message input fleet location updater acquire 2130706433 9007 1442829630625000 5601236271552241165 111825377246723855 message input fleet location updater cs 2130706433 9007 1442829630626000 5601236271552241165 111825377246723855 message input fleet location updater release 2130706433 9007 1442829635633000 3 row in set 0 16 sec this be one of 3 span in a trace where the ui report total span 3 but service location service be list twice fleet location ingest x1 fleet location updater x1 service location service x2 maybe this be a clue since the other trace not include a sr ss do not have duplicate |
openzipkinzipkin | ui query do not work if be present in the value of a binary annotation | Bug | the current split logic l76 blindly split whole of the query by and do not care if that might be present in the value hence this binnary annotation query fail http uri foo bar 3 I e have another in the value foo bar 2 |
openzipkinzipkin | allow cache of getservice in the ui | Enhancement | from yurishkuro the query to get all service name in cassandra need to hit every node in the cluster we ve be see strange huge latency with it recently still investigate to mitigate the impact I ve add a cache of service name in cassandra span storage which periodically refresh itself I think we should solve this at the http abstraction ex send cache header back so that the client doesn t call so much there be two endpoint on zipkin web that call getservice part of a massive thing that include enumerate all service to pass to index mustache api service possibly unused json string list I see two thing to do change index mustache to use javascript to get api service and fill the service name selector by itself change handleservice to return cache header |
openzipkinzipkin | reduce duplicate write in cassandra to service name and span name table | Enhancement | problem store span in cassandraspanstore involve write to service name and span name table these table be nothing more than list of service name and span name active within the index ttl period default of 3 day there be both at the application level and in the database a large amount of waste overhead with duplicate write have to do this for every span solution keep a simple thread local cache in the write path so within any give hour write to either table that have already happen be avoid this remove the need for the span name bucket column but it s keep for compatibilty and a value of zero always now use in it |
openzipkinzipkin | zipkin collector service be well configurable | Enhancement | allow more configuration option to be configurable through environment variable especially be able to easily configure log level sample rate and number of queue worker be crucial for we |
openzipkinzipkin | revisit zipkin aggregate | Duplicate | zipkin aggregate have 3 part an api a ui and a hadoop job that populate datum I ve see this in eirslett s demo and get excite run the hadoop jar and click on the ui and scratch my head we need to inventory what this do so that folk know what it be suppose to do the lack of knowledge on how to use the aggregate and the lack of implementation of aggregate lead to issue such as 634 it seem we should be able to present this feature in a way that other can implement it ideally without additional dependency we also have an opportunity to make a turn for the well the last few month have include a large share of work due to deprecate or orphan library from twitter not to mention scala versione issue for example aggregate depend on twitter library such as algebird directly but there s no reason to do that as it could be an implementation choice the hadoop job package into a 50 mb file to support very straight forward code finally there be no test and the setup be heavy I don t know anyone besides eirslett who actually have this work I d love for we to have a dependency free api that the ui use and that could be implement in a strong and less complex tool I know that during the workshop yanowitz mention their post zipkin trace infrastructure would do analysis via spark since this be a new set of individual and it be now 2015 we have a choice in how we progress it would be good to have the most supportable least orphan risk approach what might that be |
openzipkinzipkin | can not show aggregate on web ui | Question | hi I try to display some agregregate from on web ui and it doesn t work for I here be step I make vim src main scala com twitter zipkin example main scala l22 val gensampletrace flag gensampletrace false generate sample trace val gensampletrace flag gensampletrace true generate sample trace vim zipkin anormdb src main scala com twitter zipkin anormdb anormdbspanstorefactory scala l 23 val anormdb flag zipkin storage anormdb db sqlite memory name the database type val anormdb flag zipkin storage anormdb db sqlite persistent name the database type gradlew zipkin example run sqlite3 build zipkin db zipkin tracegen src testdata aggregate sql I expect that after this action some aggregate will be display at but they be not |
openzipkinzipkin | add environment app name in the web ui | Enhancement | this commit add an environment or app name in the top right corner of the the ui this argument default to the empty string so there will be no visual change to previous version if people don t want to set it |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.