owner stringclasses 14
values | repo stringclasses 14
values | id int64 116k 4.61B | issue_number int32 1 181k | author stringlengths 1 39 | body stringlengths 1 262k | created_at timestamp[us]date 2000-06-06 02:40:44 2026-06-02 20:16:18 | updated_at timestamp[us]date 2000-06-06 02:40:44 2026-06-02 20:22:56 | reactions unknown | author_association stringclasses 5
values |
|---|---|---|---|---|---|---|---|---|---|
ClickHouse | ClickHouse | 785,321,762 | 17,113 | smagellan | @alexey-milovidov, are you talking about 'Prepared Statements' concept (parametrized query AST caching on server side, etc)? | 2021-02-24T19:35:06 | 2021-02-24T19:35:06 | {} | NONE |
ClickHouse | ClickHouse | 785,425,209 | 21,089 | den-crane | probably because postpone is logarithmic. Check system.replicated_queue
| 2021-02-24T22:25:11 | 2021-02-24T22:25:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,519,869 | 21,115 | SE2AI | Thank you very much, I understand how it works now: ) | 2021-02-25T02:01:41 | 2021-02-25T02:01:41 | {} | NONE |
ClickHouse | ClickHouse | 785,710,145 | 19,580 | alesapin | > @alesapin I'll appreciate it if you fix that. Either fix the type issue, the epoll issue and possible others, or disable NuKeeper module for OS_DARWIN for now.
But it's already disabled? https://github.com/ClickHouse/ClickHouse/blob/master/src/CMakeLists.txt#L197-L199 | 2021-02-25T08:18:51 | 2021-02-25T08:18:51 | {} | MEMBER |
ClickHouse | ClickHouse | 785,763,491 | 14,614 | yusufozturk | @4ertus2 This is now working for us. But nobody changed anything or upgraded in our environment. I don't know what to do now. Should we trust this as "it's working" or we should continue to test it? | 2021-02-25T09:45:32 | 2021-02-25T09:45:32 | {} | NONE |
ClickHouse | ClickHouse | 785,004,240 | 21,135 | ngsiolei | `interserver_http_host` and `interserver_http_port` are used for data replication between servers. `interserver_http_host` defaults to the output of `hostname -f`.
Some `interserver_*` related settings and comments in `config.xml` are as follows
```
<!-- Port for communication between replicas. Used for data e... | 2021-02-24T11:16:17 | 2021-02-24T11:16:17 | {
"+1": 2
} | NONE |
ClickHouse | ClickHouse | 785,066,705 | 21,041 | OmarBazaraa | Adding more context ...
I am not using a native MongoDB server; I am using CosmosDB deployed on Azure which has MongoDB interface.
I can normally connect to my CosmosDB instance using MongoDB Shell:
```
mongo --host <host> --port <port> --username <username> --password <password> --ssl
```
And I can norma... | 2021-02-24T13:15:23 | 2021-02-24T13:15:23 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,229,930 | 20,879 | vitlibar | > It's unclear how correctly the method `read` works (if it does not set skip degree)?
It doesn't seem to work correctly, and the previous version had the same problem, it didn't set the skip degree in `read()` either. I'm not sure if it's possible to fix that without changing the format. | 2021-02-24T17:10:01 | 2021-02-24T17:23:12 | {} | MEMBER |
ClickHouse | ClickHouse | 785,293,282 | 20,879 | alexey-milovidov | Will be possible to fix after #12552 | 2021-02-24T18:47:10 | 2021-02-24T18:47:10 | {} | MEMBER |
ClickHouse | ClickHouse | 785,307,376 | 21,136 | filimonov | See also #6294 #208 | 2021-02-24T19:09:35 | 2021-02-24T19:09:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,514,318 | 18,779 | zhangjmruc | @AnaUvarova This new function POSITION(needle IN haystack) works exactly same as POSITION(haystack, needle). I don't dig in how it works, what I do is related to only some parser works to allow parser to accept the new syntax.
> Could you please help me understand how exactly this new functionality works?
> Maybe so... | 2021-02-25T01:45:39 | 2021-02-25T01:45:39 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,538,830 | 8,997 | chyy | Had the similar problem after upgrade from 20.12.5.14
[ch-01] 2021.02.24 20:31:13.845342 [ 17553 ] <Fatal> BaseDaemon: ########################################
[ch-01] 2021.02.24 20:31:13.845469 [ 17553 ] <Fatal> BaseDaemon: (version 21.2.4.6 (official build), build id: 6E2E926515F53BCF094B3D6BD7E9EB65BC5BCE6B) (fr... | 2021-02-25T02:56:00 | 2021-02-25T02:56:00 | {} | NONE |
ClickHouse | ClickHouse | 785,578,826 | 21,135 | cokepizza | However, I also confirmed that replication works normally even if that value is set strangely, but why do you think it is? Is that key all involved in the replication? For reference, our clickhouse works with zookeeper. | 2021-02-25T04:12:47 | 2021-02-25T04:13:42 | {} | NONE |
ClickHouse | ClickHouse | 785,692,043 | 20,963 | kitaisreal | @mikekytyzov we added support for single tuple argument about two month ago.
Connected to ClickHouse server version 21.3.1 revision 54447.
```
CREATE TABLE salary (user String, salary UInt64) ENGINE=TinyLog;
INSERT INTO salary VALUES ('user', 5000), ('manager', 3000), ('worker', 1000);
SELECT argMax(user, salary... | 2021-02-25T07:44:58 | 2021-02-25T07:44:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,988,867 | 21,080 | taylor12805 | > ORC & Parquet (not sure) allows Null instead of list.
> Let's implement these
>
> from ORC(list) Clickhouse back to ORC(list)
> [] [] []
> [null] [null] [null]
> null [] []
> ```python
> with open("./test4.orc", "wb") as data:
> with pyorc.Writer(data, "struct<col0:array<int>,col1:array<string>>") as wr... | 2021-02-24T10:50:37 | 2021-02-24T10:50:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,082,684 | 21,139 | den-crane | MV is an insert trigger. It gets data from insert. **IT NEVER READS THE SOURCE TABLE.**
ENGINE OF THE SOURCE TABLE DOES NOT MATTER.
Check this:
https://den-crane.github.io/Everything_you_should_know_about_materialized_views_commented.pdf
https://youtu.be/ckChUkC3Pns?list=PLO3lfQbpDVI-hyw4MyqxEk3rDHw95SzxJ&t=9371 | 2021-02-24T13:40:33 | 2021-02-24T13:40:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,088,654 | 20,087 | Felixoid | Ok, it means, that the rollup.xml works as expected. And now the issue is the fact, that from the graphite-ch-optimizer perspective, partitions aren't stale enough.
This ticket can be closed, there's nothing to do from the ClickHouse perspective. | 2021-02-24T13:50:17 | 2021-02-24T13:50:17 | {} | MEMBER |
ClickHouse | ClickHouse | 785,160,507 | 21,121 | kitaisreal | @amosbird could you please explain fix in details ? | 2021-02-24T15:33:34 | 2021-02-24T15:33:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,254,426 | 18,779 | AnaUvarova | Hi, @zhangjmruc !
Could you please help me understand how exactly this new functionality works?
Maybe some interesting details about it? | 2021-02-24T17:45:25 | 2021-02-24T17:45:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,973,191 | 21,108 | alexey-milovidov | To avoid overflow we can simply use Float64 as accumulator. It will have two caveats:
- drop in performance;
- incompatible aggregation state, that requires either to: - have a separate aggregate function; - implement #12552. | 2021-02-24T10:24:42 | 2021-02-24T10:24:42 | {} | MEMBER |
ClickHouse | ClickHouse | 784,985,091 | 20,964 | pronvis | Update to `(version 21.2.4.6 (official build))` didn't help.
---
So the last idea is to move Zookeepers away from Clickhouse nodes? | 2021-02-24T10:44:08 | 2021-02-24T10:44:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,115,082 | 21,141 | alexey-milovidov | Compact format is also column-oriented but with columns placed in one file. | 2021-02-24T14:30:23 | 2021-02-24T14:30:23 | {} | MEMBER |
ClickHouse | ClickHouse | 785,107,009 | 20,968 | vitlibar | > @vitlibar It seems that after this PR the build of utils is broken:
>
> ```
> cmake -DCMAKE_C_COMPILER=/usr/bin/clang -DCMAKE_CXX_COMPILER=/usr/bin/clang++ -DCMAKE_BUILD_TYPE=Debug -DPARALLEL_LINK_JOBS=12 -DPARALLEL_COMPILE_JOBS=56 -G Ninja ../../ClickHouse/
>
> ninja: error: '/home/jakalletti/ClickHouse/Click... | 2021-02-24T14:18:28 | 2021-02-24T16:00:44 | {} | MEMBER |
ClickHouse | ClickHouse | 785,292,114 | 21,020 | azat | > @azat Do you know why fuzzer did not stop after first segfault?
It should be because SIGSEGV handler does not call _exit, so when thread got SIGSEGV it just send information about the signal to the pipe, and the server will be terminated only once this message in pipe will be served, but this may take a while, in ... | 2021-02-24T18:45:11 | 2021-02-24T18:45:11 | {} | MEMBER |
ClickHouse | ClickHouse | 785,301,770 | 15,936 | filimonov | It seems it does not flap currently, may be it was some temp CI issue. | 2021-02-24T19:00:11 | 2021-02-24T19:00:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,348,699 | 20,963 | kitaisreal | Hello @mikekytyzov.
Documentation seems fine, what is the purpose of this change ? | 2021-02-24T20:20:52 | 2021-02-24T20:20:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,531,529 | 20,167 | yiguolei | @Jokser
If 64MB < part size < 512MB, then it will use multi part upload, but there is only a single part to upload and it will require 3 requests.
Why not just use 512MB as threshold. If part size < 512MB use single part upload and if part size > 512MB use multipart upload. | 2021-02-25T02:34:14 | 2021-02-25T02:34:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,585,764 | 10,505 | huangs0 | This might be a problem of clickhouse-client.
Actually if you installing clickhouse it's already run on the Colab
You can type following code to verify
`!lsof -i:9000`
clickhouse will use the port 9000 in tcp mode
You shall try other connector like its driver in Python
`!pip install clickhouse_driver`
Remember ... | 2021-02-25T04:16:52 | 2021-02-25T04:16:52 | {} | NONE |
ClickHouse | ClickHouse | 785,666,279 | 20,946 | tavplubix | Stress test (thread) - OOM | 2021-02-25T07:02:06 | 2021-02-25T07:02:06 | {} | MEMBER |
ClickHouse | ClickHouse | 785,709,533 | 19,580 | tisonkun | @alesapin I'll appreciate it if you fix that. Either fix the type issue, the epoll issue and possible others, or disable NuKeeper module for OS_DARWIN for now. | 2021-02-25T08:17:37 | 2021-02-25T08:17:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,067,286 | 19,291 | Avogar | @KochetovNicolai can you review he changes? | 2021-02-24T13:16:25 | 2021-02-24T13:16:25 | {} | MEMBER |
ClickHouse | ClickHouse | 785,112,133 | 19,766 | vdimir | @achimbab
I left comments, generally it looks good, but core algorithm not so clear, I left some comments, I think it need to be clarified in more details.
Also I made minor changes (mostly style) and pushed it.
| 2021-02-24T14:26:16 | 2021-02-24T14:26:23 | {} | MEMBER |
ClickHouse | ClickHouse | 785,183,069 | 11,117 | vitlibar | Fixed, the example turned into the test `00825_protobuf_format_nested_in_nested`. | 2021-02-24T16:04:40 | 2021-02-24T16:04:40 | {} | MEMBER |
ClickHouse | ClickHouse | 785,647,305 | 21,096 | fuwhu | > This is a nice feature to have. I'd suggest using `CREARE FORMAT SCHEMA ` instead, as `SCHEMA` might lead to unnecessary confusion. We can also extend those Formats to accept defined format schema names instead of using file names.
@amosbird Do you mean like this :
_SELECT * FROM test.table FORMAT Protobuf SETTI... | 2021-02-25T06:21:25 | 2021-02-25T06:21:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,973,418 | 21,108 | alexey-milovidov | Workaround: covert argument to Float64 when passing to avg. | 2021-02-24T10:25:02 | 2021-02-24T10:25:02 | {} | MEMBER |
ClickHouse | ClickHouse | 785,074,216 | 19,291 | Avogar | @alexey-milovidov I am thinking of adding new metric for PoolState: `slowdown_count`. It's like errors_count, but we will increment it when timeout for changing replica expired, so it will help to move more healthy replicas to the first places in ShuffledPool. What do you think? | 2021-02-24T13:27:34 | 2021-02-24T13:27:34 | {} | MEMBER |
ClickHouse | ClickHouse | 785,088,526 | 21,139 | den-crane | Try tuple comparison (selectivity, timestamp)
```
SELECT argMax(timestamp, (selectivity, timestamp) ) as time_at_max_selectivity,
type,
max(selectivity) as max_selectivity
FROM my_data
GROUP BY type
ORDER BY type;
```
```
argMaxState(timestamp, (selectivity, timestamp) )
``` | 2021-02-24T13:50:04 | 2021-02-24T13:50:13 | {
"rocket": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,157,872 | 21,121 | amosbird | Test failure is unrelated. | 2021-02-24T15:29:58 | 2021-02-24T15:29:58 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 785,269,285 | 20,968 | vitlibar | > ANTLR protobuf tests broken now? https://clickhouse-test-reports.s3.yandex.net/20968/2eecaee08dc1f52786527674d3e9263df846cdb1/functional_stateless_tests_(antlr_debug).html#fail1
`skip_list.json` will be fixed in https://github.com/ClickHouse/ClickHouse/pull/21148 | 2021-02-24T18:08:48 | 2021-02-24T18:39:40 | {} | MEMBER |
ClickHouse | ClickHouse | 785,515,813 | 21,095 | zhangjmruc | The 1 failed functional stateless test is not related to clickhouse-local, but related to unclean of zookeeper path.
details of failed message
-------------------
2021-02-24 15:25:50 01035_concurrent_move_partition_from_table_zookeeper: [ FAIL ] 30.36 sec. - return code 232
2021-02-24 15:25:50 ... | 2021-02-25T01:49:55 | 2021-02-25T01:50:50 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,660,187 | 20,219 | yiguolei | @qoega
> * vshard key can be misused by a user and data will be distributed with misbalance. Some shards can be overloaded and other can be almost empty. There can appear a new problem to autobalance vshards.
> * if user defines vshard key manually it can lead to data duplicates if he has incorrect code. vshards ... | 2021-02-25T06:48:55 | 2021-02-25T07:07:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,071,404 | 18,884 | WingsGo | Hi, @fuwhu @sundy-li I found that window Funnel agg function still have inconsistent result in some case I haven't found , for example,
SELECT level, count(*) AS c
FROM (
SELECT user_id
, windowFunnel(86400)(lst_vst_time_1d, spma = 'a106'
AND spmb = 'b1', spma = 'a106'
AND spmb = 'b2') AS level
FROM t... | 2021-02-24T13:23:24 | 2021-02-24T13:23:44 | {} | NONE |
ClickHouse | ClickHouse | 785,117,696 | 20,991 | alexey-milovidov | There are at least two possibilities:
- we pass uninitialized data to brotli or using its interface incorrectly;
- brotli is using uninitizlied memory in legitimate way in its algorithms and it is not annotated for memory sanitizer (unlikely). | 2021-02-24T14:34:17 | 2021-02-24T14:34:17 | {} | MEMBER |
ClickHouse | ClickHouse | 785,264,913 | 20,294 | sevirov | @kitaisreal Please, review this PR. | 2021-02-24T18:02:01 | 2021-02-24T18:02:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,304,572 | 21,136 | filimonov | Significant limitation of that is that most of http clients don't have a good support of trailers, and reimplementing that may require hacking http client on quite a low level (chunked decoding).
Does poco support that? | 2021-02-24T19:04:39 | 2021-02-24T19:04:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,529,355 | 16,702 | amosbird | Currently none of these partition related alter commands are atomic. We only have part-level atomicity. | 2021-02-25T02:28:04 | 2021-02-25T02:28:04 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 785,627,949 | 19,897 | funemy | Hi, just to update a little bit -- we do have a release candidate [here](https://public-installer-pkg.s3.us-east-2.amazonaws.com/coderrect-linux-hpc-1.0.0.tar.gz)
However, I'm not 100% sure if it can analyze clickhouse consistently. we had some successful scans before. But I'm recently having some new errors when sc... | 2021-02-25T05:40:52 | 2021-02-25T05:40:52 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 785,668,349 | 20,585 | alesapin | Ok, let's merge and see what will happen. | 2021-02-25T07:05:55 | 2021-02-25T07:05:55 | {} | MEMBER |
ClickHouse | ClickHouse | 785,690,179 | 20,963 | mikekytyzov | @kitaisreal not really fine :)
Current function accepts always only two args, but in the examples provided below it became with only one, what is not valid :) Example usage raises error. | 2021-02-25T07:41:16 | 2021-02-25T07:41:40 | {} | NONE |
ClickHouse | ClickHouse | 785,694,137 | 20,963 | mikekytyzov | @kitaisreal Okay. Sorry, i did not saw it on 21.3.1 version. I was checking on little lower. Thanks :) | 2021-02-25T07:48:57 | 2021-02-25T07:48:57 | {} | NONE |
ClickHouse | ClickHouse | 785,704,658 | 19,580 | tisonkun | > > commit [57c9b6c](https://github.com/ClickHouse/ClickHouse/commit/57c9b6c864d10909e74cd02e9ba4c90cd4487339) belongs to this pull request excludes darwin(macOS) from running ClickHouse with
> > ```
> > <Error> Application: DB::Exception: ClickHouse server built without NuRaft library. Cannot use internal coordinati... | 2021-02-25T08:08:56 | 2021-02-25T08:10:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,267,387 | 20,478 | sevirov | @tavplubix Please, review this PR. | 2021-02-24T18:05:52 | 2021-02-24T18:05:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,377,977 | 20,946 | tavplubix | f41e68d:
Stress test (address) - `alter table lc_empty_part_bug modify column s LowCardinality(String)` - `StorageMergeTree::waitForMutation`
Stress test (thread) - `SELECT 2, count() FROM system.numbers` - bad test `00600_replace_running_query`
Stress test (undefined) - `2021.02.24 01:17:42.862194 [ 3599 ] {3816e57... | 2021-02-24T20:55:10 | 2021-02-24T23:08:10 | {} | MEMBER |
ClickHouse | ClickHouse | 785,771,288 | 19,896 | nikitamikhaylov | @depressed-pho Please, run a script `generate-ya-make` from `utils` folder to solve an issue with `Yandex synchronization check`. Then commit add all ya.make files to this PR. | 2021-02-25T09:58:14 | 2021-02-25T09:58:14 | {} | MEMBER |
ClickHouse | ClickHouse | 784,984,675 | 21,020 | alexey-milovidov | https://github.com/ClickHouse/ClickHouse/issues/21137 | 2021-02-24T10:43:33 | 2021-02-24T10:43:33 | {} | MEMBER |
ClickHouse | ClickHouse | 784,985,177 | 21,020 | alexey-milovidov | @azat Do you know why fuzzer did not stop after first segfault? | 2021-02-24T10:44:17 | 2021-02-24T10:44:17 | {} | MEMBER |
ClickHouse | ClickHouse | 785,086,266 | 20,968 | nikitamikhaylov | @vitlibar It seems that after this PR the build of utils is broken:
```
cmake -DCMAKE_C_COMPILER=/usr/bin/clang -DCMAKE_CXX_COMPILER=/usr/bin/clang++ -DCMAKE_BUILD_TYPE=Debug -DPARALLEL_LINK_JOBS=12 -DPARALLEL_COMPILE_JOBS=56 -G Ninja ../../ClickHouse/
ninja: error: '/home/jakalletti/ClickHouse/ClickHouse/tests/q... | 2021-02-24T13:46:41 | 2021-02-24T13:46:41 | {} | MEMBER |
ClickHouse | ClickHouse | 785,103,064 | 20,991 | kssenii | What if with the given arguments passed to method, where uninitialised data access is done, it is impossible to correctly define up to which position array is initialized (to add a check)? (My attempts fix msan, but some rows are lost). And Brotli always initializes less than it allocates for that array, but does it wi... | 2021-02-24T14:12:14 | 2021-02-24T14:12:14 | {} | MEMBER |
ClickHouse | ClickHouse | 785,201,654 | 21,121 | kitaisreal | @amosbird thanks for details. Make sense. | 2021-02-24T16:31:04 | 2021-02-24T16:31:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,706,954 | 19,580 | alesapin | > > > commit [57c9b6c](https://github.com/ClickHouse/ClickHouse/commit/57c9b6c864d10909e74cd02e9ba4c90cd4487339) belongs to this pull request excludes darwin(macOS) from running ClickHouse with
> > > ```
> > > <Error> Application: DB::Exception: ClickHouse server built without NuRaft library. Cannot use internal coor... | 2021-02-25T08:12:38 | 2021-02-25T08:12:38 | {} | MEMBER |
ClickHouse | ClickHouse | 785,758,979 | 21,009 | vdimir | @KochetovNicolai
> `std::shared_lock<std::shared_mutex> storage_join_lock` is not needed in `HashJoin`. `HashJoin` should not be responsible for external synchronization.
We discussed that it will requires changes in ExpressionAnalyzer, it will have to handle this lock. So decided to keep lock in HashJoin and ad... | 2021-02-25T09:38:42 | 2021-02-25T09:39:01 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 784,989,446 | 20,167 | Jokser | @yiguolei
Single part upload requires 1 request to S3 and useful for small uploads.
Multipart upload requires at least 3 requests to S3:
initiate multipart upload
upload part 1
upload part 2
...
upload part N
complete multipart upload
Multipart upload is needed to upload large files that need to be split ... | 2021-02-24T10:51:42 | 2021-02-24T10:51:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,668,110 | 20,585 | alesapin | third-party -- not related
ANTLR debug -- not related, broken in https://github.com/ClickHouse/ClickHouse/pull/20968 | 2021-02-25T07:05:27 | 2021-02-25T07:05:27 | {} | MEMBER |
ClickHouse | ClickHouse | 785,692,620 | 21,060 | mxzlxy | @tavplubix Functional stateless tests still failed, how can I fix it, thanks | 2021-02-25T07:46:08 | 2021-02-25T07:46:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,061,594 | 16,979 | vitlibar | Serialization of complex structures in the Protobuf format has been fixed, it must work now. | 2021-02-24T13:06:28 | 2021-02-24T13:06:28 | {} | MEMBER |
ClickHouse | ClickHouse | 785,061,644 | 21,141 | UnamedRus | You need to disable compact format
```
ALTER TABLE test.order MODIFY SETTING min_bytes_for_wide_part=0,min_rows_for_wide_part=0
``` | 2021-02-24T13:06:33 | 2021-02-24T13:06:33 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,251,887 | 21,009 | KochetovNicolai | Some considerations:
* `std::shared_lock<std::shared_mutex> storage_join_lock` is not needed in `HashJoin`. `HashJoin` should not be responsible for external synchronization.
* `JoinSource` is an internal class for `HashJoin`. It need to get `HashJoinPtr` + `std::shared_lock<std::shared_mutex> storage_join_lock` in ... | 2021-02-24T17:41:38 | 2021-02-24T17:41:38 | {} | MEMBER |
ClickHouse | ClickHouse | 785,335,821 | 19,291 | KochetovNicolai | Everything else LGTM in general. (I did not check tests properly so far). | 2021-02-24T20:00:03 | 2021-02-24T20:00:03 | {} | MEMBER |
ClickHouse | ClickHouse | 785,353,609 | 21,020 | azat | >But they are from different queries (check query_id) and fuzzer is sending queries sequentially.
Yep, but the problem is that the first query failed with different error - `Data compressed with different methods`:
```
Error on processing query 'SELECT item_id FROM (SELECT item_id FROM t GROUP BY item_id WITH TO... | 2021-02-24T20:28:31 | 2021-02-24T20:28:31 | {} | MEMBER |
ClickHouse | ClickHouse | 785,652,479 | 21,096 | amosbird | > Am i right ?
exactly | 2021-02-25T06:32:33 | 2021-02-25T06:32:33 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 785,716,685 | 21,135 | ngsiolei | I did a simple test to explicitly set a wrong value of `interserver_http_host`, got 'Host not found' error and failed in data replication. Wrong `interserver_http_host` is likely one of the causes introduce 'Host not found' error, your case may be due to other causes, just see this test as a reference.
## test deta... | 2021-02-25T08:31:02 | 2021-02-25T08:31:02 | {} | NONE |
ClickHouse | ClickHouse | 784,965,019 | 21,128 | alexey-milovidov | Run `ninja clickhouse` instead of `ninja`. | 2021-02-24T10:12:00 | 2021-02-24T10:12:00 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 785,060,636 | 19,291 | Avogar | > Please check if this fixes #9900.
Yes, it fixes. I will add such test. | 2021-02-24T13:04:48 | 2021-02-24T13:04:48 | {} | MEMBER |
ClickHouse | ClickHouse | 785,179,503 | 21,121 | amosbird | Setting `input_format_null_as_default` should be false when the target column type is Nullable. https://github.com/ClickHouse/ClickHouse/blob/master/src/Core/Settings.h#L419
It is set to true by default in https://github.com/ClickHouse/ClickHouse/pull/17525 , and somehow we turn it on for Nullable types in `Constan... | 2021-02-24T15:59:54 | 2021-02-24T16:00:35 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 785,257,268 | 20,968 | alesapin | ANTLR protobuf tests broken now? https://clickhouse-test-reports.s3.yandex.net/20968/2eecaee08dc1f52786527674d3e9263df846cdb1/functional_stateless_tests_(antlr_debug).html#fail1 | 2021-02-24T17:49:54 | 2021-02-24T17:49:54 | {} | MEMBER |
ClickHouse | ClickHouse | 785,704,150 | 15,090 | alesapin | More related PRs:
https://github.com/ClickHouse/ClickHouse/pull/19580
https://github.com/ClickHouse/ClickHouse/pull/20585
| 2021-02-25T08:08:07 | 2021-02-25T08:08:07 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 785,058,057 | 7,438 | vitlibar | The constants of a [Enum data type](https://clickhouse.tech/docs/en/sql-reference/data-types/enum/) are now matched to constants of a [protobuf enum](https://developers.google.com/protocol-buffers/docs/proto3?hl=ru#enum) first by name (case insensitively), then by value. So this example now works, and test was added (s... | 2021-02-24T13:00:14 | 2021-02-24T13:00:22 | {} | MEMBER |
ClickHouse | ClickHouse | 785,080,989 | 21,115 | den-crane | >So If active equals to 1 for some partitions, it means I need not to care
>about how many partitions there are, right? Even there are some single partitions?
right.
>If merge is resource expensive, what is OPTIMIZE designed for and
> what are the most appropriate application scenarios for using it?
1. debu... | 2021-02-24T13:37:45 | 2021-02-24T13:38:25 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,143,933 | 2,531 | filimonov | Fixed long ago, test added in https://github.com/ClickHouse/ClickHouse/pull/21111 | 2021-02-24T15:11:39 | 2021-02-24T15:11:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,299,896 | 21,020 | alexey-milovidov | But they are from different queries (check query_id) and fuzzer is sending queries sequentially. | 2021-02-24T18:57:09 | 2021-02-24T18:57:22 | {} | MEMBER |
ClickHouse | ClickHouse | 785,361,007 | 21,089 | cw9 | Why do I sometimes see no replicated fetches
```
SELECT count()
FROM system.replicated_fetches
Query id: 403d71a7-06e8-430a-a152-a94ce1843321
┌─count()─┐
│ 0 │
└─────────┘
```
When there are tables data still to be replicated?
```
SELECT count()
FROM system.replicas
WHERE queue_size > 20
Qu... | 2021-02-24T20:41:18 | 2021-02-24T20:41:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,285,938 | 20,946 | tavplubix | Stress test (address) - `ALTER TABLE mt_00763_1 MODIFY column s UInt32` hung waiting for mutation - `StorageMergeTree::waitForMutation`
Stress test (thread) - `WATCH test.lv EVENTS` - #21004
Stress test (undefined) - `ALTER TABLE collapsing_00682 DELETE WHERE k IN (1, 2)` and `alter table lc_empty_part_bug modify co... | 2021-02-23T15:31:38 | 2021-02-24T23:08:49 | {} | MEMBER |
ClickHouse | ClickHouse | 785,579,154 | 21,167 | den-crane | @chyy can you grep your logs for b03fb6ff-7fe1-400e-bb55-d71b0c611525
zgrep b03fb6ff-7fe1-400e-bb55-d71b0c611525 /var/log/clickhouse-server/* | 2021-02-25T04:12:59 | 2021-02-25T04:13:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,713,745 | 19,580 | tisonkun | https://github.com/ClickHouse/ClickHouse/blob/36c178c0231a073fdb3f75ca52a85ba9e4fbea7e/programs/server/Server.cpp#L872-L874
You're right. I should use the term "run", normally running on OS_DARWIN. However, if it is an significant burden over prototyping, it's ok we give another pass later for the fix. | 2021-02-25T08:25:42 | 2021-02-25T08:25:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,775,340 | 20,167 | Jokser | @yiguolei
Because it's AWS recommendation:
> Multipart upload allows you to upload a single object as a set of parts. Each part is a contiguous portion of the object's data. You can upload these object parts independently and in any order. If transmission of any part fails, you can retransmit that part without aff... | 2021-02-25T10:04:16 | 2021-02-25T10:04:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,968,730 | 21,058 | kitaisreal | @windylcx try to reinstall llvm and clang in homebrew. | 2021-02-24T10:17:33 | 2021-02-24T10:17:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,055,363 | 9,069 | vitlibar | Fixed, the example turned into the test `00825_protobuf_format_array_of_arrays`. | 2021-02-24T12:55:16 | 2021-02-24T13:03:06 | {} | MEMBER |
ClickHouse | ClickHouse | 785,308,352 | 21,136 | alexey-milovidov | > Does poco support that?
No.
But we can send trailers only if client requested them:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/TE | 2021-02-24T19:11:17 | 2021-02-24T19:11:17 | {} | MEMBER |
ClickHouse | ClickHouse | 784,971,469 | 21,128 | liuzhizu | thank you @alexey-milovidov | 2021-02-24T10:21:54 | 2021-02-24T10:21:54 | {} | NONE |
ClickHouse | ClickHouse | 784,971,998 | 21,108 | alexey-milovidov | @UnamedRus Need to update documentation to describe this behaviour. | 2021-02-24T10:22:45 | 2021-02-24T10:22:45 | {} | MEMBER |
ClickHouse | ClickHouse | 785,059,132 | 6,497 | vitlibar | Fixed, the example turned into the test `00825_protobuf_format_nested_optional`. | 2021-02-24T13:02:07 | 2021-02-24T13:02:07 | {} | MEMBER |
ClickHouse | ClickHouse | 785,066,488 | 21,108 | UnamedRus | BigInt is actually faster and provide more correct result than Float64.
```
set allow_experimental_bigint_types=1;
SELECT avg(toInt128(number))
FROM numbers(4294967300)
Query id: 4e116c13-483c-423c-a702-0b69825351de
┌─avg(toInt128(number))─┐
│ 2147483649.5 │
└───────────────────────┘
1 rows ... | 2021-02-24T13:15:02 | 2021-02-24T13:15:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,101,774 | 21,139 | cyber-moon | Hey @den-crane
Thanks for the fast answer! I wasn't aware that we can maximize over tuples, this definitely solves the problem and is way nicer than my approach. Thanks a lot! | 2021-02-24T14:10:16 | 2021-02-24T14:10:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,109,042 | 21,029 | kitaisreal | Fixed by https://github.com/ClickHouse/ClickHouse/pull/21120. | 2021-02-24T14:21:33 | 2021-02-24T14:21:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,183,811 | 21,009 | vdimir | `test_storage_kafka/test.py::test_kafka_no_holes_when_write_suffix_failed` not related, failed in master during past three days | 2021-02-24T16:05:39 | 2021-02-24T16:05:39 | {} | MEMBER |
ClickHouse | ClickHouse | 785,313,740 | 21,136 | filimonov | My IMHO: trailers are nice because they are part of http itself, and that's the only advantage. The main disadvantage is how they supported by http clients. For ex. the most natural http client (browser) can't work with them.
Server-sent-events are simpler to implement, work great in browsers, has nice libs to work... | 2021-02-24T19:20:24 | 2021-02-24T19:29:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 785,716,220 | 21,122 | kitaisreal | Hello, @Slach. Please remove unnecessary changes in contrib. Currently there is +34 −291,178 diff. | 2021-02-25T08:30:17 | 2021-02-25T08:30:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 792,280,697 | 21,503 | den-crane | This result is from JDBC application? Right?
Can you show the result of
```
select timezone(), now(), toString(now());
``` | 2021-03-07T13:35:56 | 2021-03-07T13:35:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 792,289,417 | 21,087 | vzakaznikov | > Writing SHALL in requirements to just document current behaviour looks wrong.
> I recomment to replace it with MAY.
Agree. I will update the requirements that just document current behavior to use MAY instead of SHALL. | 2021-03-07T14:38:40 | 2021-03-07T14:39:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 792,289,699 | 21,425 | alesapin | No idea how is it possible. I'll upload of coordination dump after stress test. | 2021-03-07T14:40:50 | 2021-03-07T14:40:50 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.