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 | 654,339,168 | 8,706 | alexey-milovidov | > Maybe we can check the validity of the parsed date like we just did in parseDateTimeBestEffort?
Yes, but we need to carefully check how it will affect performance. | 2020-07-06T16:27:43 | 2020-07-06T16:27:43 | {} | MEMBER |
ClickHouse | ClickHouse | 654,347,565 | 4,007 | alexey-milovidov | `__swap_writepage`
You have to disable swap and/or increase memory amount on your server.
https://clickhouse.tech/docs/en/operations/tips/ | 2020-07-06T16:44:32 | 2020-07-06T16:44:32 | {} | MEMBER |
ClickHouse | ClickHouse | 654,497,859 | 12,142 | akuzm | Let's merge and look how it goes, we can roll it back if it's too bad. | 2020-07-06T22:34:36 | 2020-07-06T22:34:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,038,826 | 12,158 | alexey-milovidov | @tavplubix Don't understand why someone removes the `.sql.tmp` file. | 2020-07-06T06:21:02 | 2020-07-06T06:21:02 | {} | MEMBER |
ClickHouse | ClickHouse | 654,038,900 | 12,147 | alexey-milovidov | Stress test: #12158 | 2020-07-06T06:21:16 | 2020-07-06T06:21:16 | {} | MEMBER |
ClickHouse | ClickHouse | 654,267,186 | 11,942 | filimonov | Probably we can refresh setting inside ddl worker somehow. | 2020-07-06T14:17:42 | 2020-07-06T14:17:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,283,364 | 7,772 | alexey-milovidov | @Akazz is working on this issue (or perhaps not). | 2020-07-06T14:46:38 | 2020-07-06T14:47:07 | {} | MEMBER |
ClickHouse | ClickHouse | 654,294,479 | 11,795 | genzgd | How does this affect zookeeper usage and performance? Currently we have ClickHouse in multiple datacenters, but we use the `replicated_can_become_leader` to ensure all leaders are in the same datacenter as the "voting" zookeeper cluster. Before we did that we saw issues with too many active merges. Is the `replicate... | 2020-07-06T15:05:34 | 2020-07-06T15:05:34 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,301,188 | 7,124 | filimonov | Can't reproduce:
```
create table test_01 engine=MergeTree ORDER BY number AS SELECT * FROM numbers(1000);
create table test_01_d engine=MergeTree ORDER BY number AS SELECT * FROM numbers(900);
alter table test_01 delete where number in (select number from test_01_d);
SELECT count()
FROM test_01
┌─count(... | 2020-07-06T15:17:38 | 2020-07-06T15:17:38 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,343,396 | 11,795 | alexey-milovidov | > Is the replicated_can_become_leader setting deprecated?
The setting remains and has the same effect.
We allow multiple leaders but you can disable leadership for specific replicas with `replicated_can_become_leader`.
> Also, do we have to upgrade all replicas to 20.5 at the same time?
No, it's 100% compatib... | 2020-07-06T16:36:11 | 2020-07-06T16:36:11 | {} | MEMBER |
ClickHouse | ClickHouse | 654,449,225 | 12,124 | alexey-milovidov | Ok.
What we need to make "bare" Docker container full featured?
- add ca-certificates;
- as it supports custom entrypoint.sh, we need to add bash or dash or whatever...
- from this point, things become more complicated. | 2020-07-06T20:29:09 | 2020-07-06T20:29:09 | {} | MEMBER |
ClickHouse | ClickHouse | 654,462,124 | 12,127 | thyn | Another Test. Nullable before SimpleAggregateFunction.
```
CREATE TABLE temp.test7 (
d Date,
f1 Nullable(SimpleAggregateFunction(anyLast, UInt8)),
f2 Nullable(SimpleAggregateFunction(anyLast, UInt8))
)
ENGINE = AggregatingMergeTree()
PARTITION BY tuple()
ORDER BY d
```
II see that anyLast i... | 2020-07-06T20:59:08 | 2020-07-06T20:59:08 | {} | NONE |
ClickHouse | ClickHouse | 653,996,771 | 12,041 | alexey-milovidov | I'm only worried in WARNING messages during configuration stage. We have them for more than a week.
https://github.com/ClickHouse/ClickHouse/pull/11300#discussion_r446620461
As long as these messages are gone, I'm Ok with merging this PR.
@abyss7 You can continue in another PR. | 2020-07-06T03:31:43 | 2020-07-06T03:32:03 | {} | MEMBER |
ClickHouse | ClickHouse | 653,997,829 | 11,815 | alexey-milovidov | Integration test:
test_distributed_ddl/test.py::test_rename[configs_secure]
https://clickhouse-test-reports.s3.yandex.net/0/5d2013f2240f450e25323a482e3a02312dc92849/integration_tests_(thread).html#fail1 | 2020-07-06T03:36:34 | 2020-07-06T03:36:34 | {} | MEMBER |
ClickHouse | ClickHouse | 654,204,747 | 12,158 | tavplubix | Possible reasons:
- Exception from [`reloadDictionaryConfig(...)`](https://github.com/ClickHouse/ClickHouse/blob/99e9b15e00695cbb8532b56cf4fd57c271cddccf/src/Databases/DatabaseWithDictionaries.cpp#L181) after the file was successfully renamed
- Race with [`DROP DATABASE`](https://github.com/ClickHouse/ClickHouse/blob... | 2020-07-06T12:29:55 | 2020-07-06T12:29:55 | {} | MEMBER |
ClickHouse | ClickHouse | 654,210,233 | 6,638 | filimonov | Works properly since 20.1 | 2020-07-06T12:41:29 | 2020-07-06T12:41:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,346,988 | 4,007 | alexey-milovidov | @filimonov The hung happens inside Linux kernel and there is a stack trace from the kernel.
The stack trace from ClickHouse will be irrelevant.
| 2020-07-06T16:43:21 | 2020-07-06T16:43:32 | {} | MEMBER |
ClickHouse | ClickHouse | 654,421,873 | 11,469 | azat | >The bug was introduced in #10342
Interesting, but this PR does not touches the dictionary handling, looking | 2020-07-06T19:25:30 | 2020-07-06T19:25:30 | {} | MEMBER |
ClickHouse | ClickHouse | 654,504,611 | 12,142 | alexey-milovidov | Also there was an idea to sample allocations and deallocations based on hash of address (hash can be calculated in 3 clock ticks) (and do sampling with higher probability for larger sizes), but it has a downside that a loop with alloc/free will likely to always return the same address and some random loops will slowdow... | 2020-07-06T22:58:00 | 2020-07-06T23:01:01 | {} | MEMBER |
ClickHouse | ClickHouse | 653,992,666 | 8,159 | gubinjie | @wbigdata
其实上面英文部分已经讨论出了解决方案，你没有仔细看。 | 2020-07-06T03:14:02 | 2020-07-06T03:14:02 | {} | NONE |
ClickHouse | ClickHouse | 654,098,411 | 12,160 | azat | Cc @abyss7 | 2020-07-06T08:40:05 | 2020-07-06T08:40:05 | {} | MEMBER |
ClickHouse | ClickHouse | 654,123,178 | 12,142 | akuzm | ERROR: type should be string, got "https://clickhouse-test-reports.s3.yandex.net/12142/307c3c92a586e0cde9f5522607bb5d951df32103/performance_comparison/report.html#3.15\r\n\r\n500% expected variability in query time is something new... I've only seen it go to 150% before. " | 2020-07-06T09:29:22 | 2020-07-06T09:29:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,153,256 | 12,153 | alesapin | `ClickHouse build check — 19/20 builds are OK` -- docker prepare failed. | 2020-07-06T10:33:29 | 2020-07-06T10:33:29 | {} | MEMBER |
ClickHouse | ClickHouse | 654,278,905 | 11,885 | akuzm | A slightly different variant:
```
SELECT JSONExtract(JSONExtractFloat('{"a": "hello", "b": [-\000, 200.0, 300]}', 'b', 2), '{"a": "hello", "b": [-100, 200.0, 300]}', arrayJoin(['', '__.']), JSONType('{"a": "hello", "b": [-100, 200.0, 300]}', 'b'))
[akuzm-nix] 2020.07.06 17:36:03.555388 [ 1748487 ] <Fatal> BaseDa... | 2020-07-06T14:38:43 | 2020-07-06T14:38:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,281,916 | 11,235 | filimonov | https://github.com/ClickHouse/ClickHouse/issues/3708 | 2020-07-06T14:43:58 | 2020-07-06T14:43:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,468,831 | 12,127 | filimonov | i've meant `SimpleAggregateFunction(anyLast, Nullable(UInt8))` | 2020-07-06T21:14:09 | 2020-07-06T21:14:09 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,485,736 | 12,178 | azat | >No objections, let's upgrade.
https://github.com/ClickHouse/ClickHouse/pull/12181
>And also the author of Arrow has said that it does not need flatbuffers anymore.
Actually they requires flatbuffers, but do not requires flatc compiler, since it is shipped with pre-generated headers. | 2020-07-06T21:56:39 | 2020-07-06T21:56:39 | {} | MEMBER |
ClickHouse | ClickHouse | 653,997,012 | 12,041 | alexey-milovidov | @azat Is it possible to configure CMake, so these WARNING messages will be fatal errors and build will not work? | 2020-07-06T03:32:49 | 2020-07-06T03:32:49 | {} | MEMBER |
ClickHouse | ClickHouse | 654,283,084 | 5,926 | filimonov | Old issue. Can't reproduce nor on 19.9.2.4 not on newer versions.
Flee free to reopen or create new issue if you will hit that on newer ClickHouse versions. | 2020-07-06T14:46:07 | 2020-07-06T14:46:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,441,914 | 12,127 | filimonov | > anyLastIf in select is not working because it produces default value which rewrites current value.
Hm... With nulls - too? | 2020-07-06T20:11:59 | 2020-07-06T20:11:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,442,479 | 12,171 | filimonov | Will it work w/o debug symbols installed? | 2020-07-06T20:13:20 | 2020-07-06T20:13:20 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 653,992,044 | 9,621 | alexey-milovidov | We have implemented a patch for backward compatibility that leads to some performance decrease: https://github.com/ClickHouse/ClickHouse/pull/10889
It is needed for rolling upgrade when different server versions work on the same cluster.
We will remove this patch after the oldest supported version that requires it ... | 2020-07-06T03:11:02 | 2020-07-06T03:11:40 | {} | MEMBER |
ClickHouse | ClickHouse | 653,994,283 | 8,159 | wbigdata | > @wbigdata
> 其实上面英文部分已经讨论出了解决方案，你没有仔细看。
确实是，问题很大，不然周五就可以了，唉 | 2020-07-06T03:21:25 | 2020-07-06T03:21:25 | {} | NONE |
ClickHouse | ClickHouse | 654,080,845 | 9,088 | filimonov | Are you sure? :) In the recent versions, it's correct... But at the moment of posting it wasn't. And still is not correct in supported 20.1 & 20.3 versions:
```
Connected to ClickHouse server version 20.3.12 revision 54433.
0d2c584f12cc :) SELECT arrayJoin([0, 1, 3, NULL]) AS x, x = 0, if(x = 0, 'x=0', 'x<>0')... | 2020-07-06T08:01:41 | 2020-07-06T08:06:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 653,740,923 | 12,127 | thyn | It also can help to aggregate data from different sources. For example:
I have userid , userName and actioncount from one source (source 1) and userid, revenue from other source (source 2). With such aggregation I can aggregate separately data by materialized view from those sources | 2020-07-04T08:54:25 | 2020-07-06T08:37:42 | {} | NONE |
ClickHouse | ClickHouse | 654,256,076 | 8,927 | filimonov | It is still actual - 20.5.2.7, current master (20.6.1). | 2020-07-06T13:58:30 | 2020-07-06T14:00:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,268,258 | 11,933 | filimonov | the bigger number of partitions you have, the bigger zookeeper transaction it tries to start. | 2020-07-06T14:19:34 | 2020-07-06T14:19:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,441,959 | 12,179 | alexey-milovidov | But line 349 still contains a bug, the invocation of `isInjective` with bogus argument is illegal. | 2020-07-06T20:12:05 | 2020-07-06T20:12:05 | {} | MEMBER |
ClickHouse | ClickHouse | 654,442,383 | 12,180 | alexey-milovidov | Never. | 2020-07-06T20:13:05 | 2020-07-06T20:13:05 | {
"laugh": 1
} | MEMBER |
ClickHouse | ClickHouse | 654,471,641 | 11,807 | filimonov | Backport to 20.3 is needed | 2020-07-06T21:20:59 | 2020-07-06T21:20:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,476,909 | 12,142 | azat | > So we can merge :)
I got what I need (maybe I will try smaller allocations but that's it), so the only question should this be merged or not, but as @akuzm said this does not looks good:
>500% expected variability in query time is something new... I've only seen it go to 150% before.
@alexey-milovidov @akuzm... | 2020-07-06T21:33:41 | 2020-07-06T21:33:41 | {} | MEMBER |
ClickHouse | ClickHouse | 654,482,293 | 12,179 | alexey-milovidov | @azat If this method is no longer used with meaningful Block argument, we can remove this argument from the method and remove all unused implementations. | 2020-07-06T21:47:19 | 2020-07-06T21:47:19 | {} | MEMBER |
ClickHouse | ClickHouse | 654,495,168 | 12,181 | azat | Cc: @proller | 2020-07-06T22:25:28 | 2020-07-06T22:25:28 | {} | MEMBER |
ClickHouse | ClickHouse | 654,030,485 | 11,815 | alexey-milovidov | Build check still fails sporadically: https://clickhouse-builds.s3.yandex.net/0/f08ff9010eb8bd8628cf009c5d6543502432ed41/clickhouse_build_check/build_log_716952857_1594003292.txt | 2020-07-06T05:56:01 | 2020-07-06T05:56:01 | {} | MEMBER |
ClickHouse | ClickHouse | 654,234,324 | 12,127 | filimonov | Did you try inserting `anyLastIf()` ?
BTW: https://github.com/ClickHouse/ClickHouse/issues/10754#issuecomment-625891916 | 2020-07-06T13:22:45 | 2020-07-06T13:22:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,282,749 | 12,171 | alexey-milovidov | Could you please provide an example of how does it look? | 2020-07-06T14:45:30 | 2020-07-06T14:45:30 | {} | MEMBER |
ClickHouse | ClickHouse | 654,471,907 | 12,127 | thyn | I've performed 2 tests. Please, review first.
https://github.com/ClickHouse/ClickHouse/issues/12127#issuecomment-654453942
| 2020-07-06T21:21:35 | 2020-07-06T21:22:19 | {} | NONE |
ClickHouse | ClickHouse | 654,505,076 | 12,142 | alexey-milovidov | About alignment by 1 GiB... it seems totally pointless, because you can just do some allocations to return an address inside specific range of virtual address space (you can create separate arena for this purpose). | 2020-07-06T22:59:45 | 2020-07-06T22:59:45 | {} | MEMBER |
ClickHouse | ClickHouse | 654,093,877 | 12,155 | qoega | hardware/network setup is also relevant. 756MB should be sent to initiator and to client.
You can check if there is difference if you wrap it into simple aggregation with almost no output like SUM/MAX. This can show if data transfer is slower than aggregation in your case | 2020-07-06T08:30:06 | 2020-07-06T08:30:06 | {} | MEMBER |
ClickHouse | ClickHouse | 654,094,408 | 12,041 | azat | >@azat Is it possible to configure CMake, so these WARNING messages will be fatal errors and build will not work?
I thought about it, but it wasn't that easy, although possible - #12160 | 2020-07-06T08:31:16 | 2020-07-06T08:31:16 | {} | MEMBER |
ClickHouse | ClickHouse | 654,289,193 | 8,927 | alexey-milovidov | New versions work correctly (as documented), the old version (19.17) not. | 2020-07-06T14:56:35 | 2020-07-06T14:56:35 | {} | MEMBER |
ClickHouse | ClickHouse | 654,073,175 | 12,155 | filimonov | Please show the query and schema. | 2020-07-06T07:45:58 | 2020-07-06T07:45:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,139,511 | 12,142 | azat | >https://twitter.com/trav_downs/status/1262058940831092738
>A nice trick they use in another sampling memory profiler to record deallocation for the same blocks.
Interesting, but if understand it right, it will work only for "large enough" allocations | 2020-07-06T10:03:05 | 2020-07-06T10:03:05 | {} | MEMBER |
ClickHouse | ClickHouse | 654,297,382 | 11,795 | genzgd | Also, do we have to upgrade all replicas to 20.5 at the same time? | 2020-07-06T15:10:48 | 2020-07-06T15:10:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,454,606 | 12,178 | azat | Right now the problem is double-conversion, it is included in arrow with a slash, I've removed this patch.
FWIW arrow since 0.16 has double-conversion included (right now the version is 0.15, latest is 0.17), any objections on just upgrading it to 0.17? (in a separate PR of course) | 2020-07-06T20:41:27 | 2020-07-06T20:41:27 | {} | MEMBER |
ClickHouse | ClickHouse | 653,992,901 | 11,349 | alexey-milovidov | It was not fixed: https://clickhouse-test-reports.s3.yandex.net/0/ad53ef2c14da180fd56efa6439307e4650052a3d/functional_stateless_tests_(release,_databaseatomic)/test_run.txt.out.log | 2020-07-06T03:15:07 | 2020-07-06T03:15:07 | {} | MEMBER |
ClickHouse | ClickHouse | 654,155,017 | 12,142 | akuzm | Not sure, didn't study it in detail yet, but they only do this alignment for a small percentage of allocations they randomly chose to sample, so maybe it's ok...
By the way, I have another interesting thing for you :) It is viable to record every memory allocation for a query in CH, it gives about 5 times slowdown. ... | 2020-07-06T10:37:50 | 2020-07-06T11:26:30 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,282,306 | 12,156 | Slach | @blinkov thanks a lot for suggestions, please look PR again | 2020-07-06T14:44:44 | 2020-07-06T14:44:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,286,042 | 11,662 | nikitamikhaylov | Build failure is caused by internal CI problem. | 2020-07-06T14:51:26 | 2020-07-06T14:51:26 | {} | MEMBER |
ClickHouse | ClickHouse | 654,310,181 | 8,706 | ucasfl | Maybe we can check the validity of the parsed date like we just did in `parseDateTimeBestEffort`?
BTW, how to set `date_time_input_format` when insert `DateTime` data? @alexey-milovidov | 2020-07-06T15:33:33 | 2020-07-06T15:33:33 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 654,346,177 | 4,007 | alexey-milovidov | > task ParalInputsProc:14835 blocked for more than 120 seconds
Yes, it is OS/hardware issue.
It may happen when you have very high IO load and slow disk subsystem and large number of dirty pages.
Also it may happen due to issues with hardware or its drivers. | 2020-07-06T16:41:39 | 2020-07-06T16:41:39 | {} | MEMBER |
ClickHouse | ClickHouse | 653,984,160 | 8,159 | wbigdata | >3、然后你在用hdfs引擎的时候把nameNode地址写成hdfs-client.xml里面NN节点的别名，并且不要加端口号就可以了
这个地方不是写成hdfs://集群名称 这样吗？
| 2020-07-06T02:34:44 | 2020-07-06T02:34:44 | {} | NONE |
ClickHouse | ClickHouse | 654,098,708 | 6,847 | alesapin | Ok, release tests passed except `02000_optimize_skip_unused_shards_type_mismatch`. It's ok. | 2020-07-06T08:40:46 | 2020-07-06T08:40:46 | {} | MEMBER |
ClickHouse | ClickHouse | 654,148,916 | 12,142 | azat | >So it works for allocation of any size, it just has to be aligned.
It uses just an array for all available pointers of aligned allocations ((1<<47)/(1<<30)), sure this can be converted to hashtable
But (correct me if I'm wrong):
- if the alignment will be too small it will require too much memory and has too ... | 2020-07-06T10:23:03 | 2020-07-06T10:23:03 | {} | MEMBER |
ClickHouse | ClickHouse | 654,266,487 | 11,942 | filimonov | Profile settings are applied at the moment of server restart.
And yes, that it is not intuitive. :\ | 2020-07-06T14:16:36 | 2020-07-06T14:16:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,359,903 | 11,795 | genzgd | Thanks for the details! | 2020-07-06T17:08:38 | 2020-07-06T17:08:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,482,830 | 12,179 | azat | >@azat If this method is no longer used with meaningful Block argument, we can remove this argument from the method and remove all unused implementations.
I wish this was true, but it is required for dictGet* functions (since it needs attribute name to detect injectiveness) | 2020-07-06T21:48:56 | 2020-07-06T21:48:56 | {} | MEMBER |
ClickHouse | ClickHouse | 653,986,785 | 8,159 | wbigdata | > @wbigdata
> you need add "export LIBHDFS3_CONF=/etc/clickhouse-server/hdfs-client.xml " to /etc/init.d/clickhouse-server;
> 1、你把搭建hadoop的时候的hdfs-client.xml 的文件复制一份到装有ClickHouse的服务器的/etc/clickhouse-server文件夹里面。
> 2、然后在装有ClickHouse的服务器的/etc/init.d/clickhouse-server的文件里面加一行export LIBHDFS3_CONF=/etc/clickhouse-server/... | 2020-07-06T02:47:31 | 2020-07-06T02:47:31 | {} | NONE |
ClickHouse | ClickHouse | 654,286,261 | 11,933 | den-crane | >Len error 1157648
Bigger than 1MB (ZK 1MB default for jute.maxbuffer https://zookeeper.apache.org/doc/r3.3.2/zookeeperAdmin.html)
CH supports bigger buffer https://github.com/ClickHouse/ClickHouse/pull/4398
You can set bigger value in your ZK -Djute.maxbuffer= | 2020-07-06T14:51:49 | 2020-07-06T14:51:49 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,461,468 | 12,179 | azat | >But line 349 still contains a bug, the invocation of isInjective with bogus argument is illegal.
Interesting, and Block is required only for external dictionaries...
So right now it is "legal", but unclear and bad | 2020-07-06T20:57:37 | 2020-07-06T20:57:37 | {} | MEMBER |
ClickHouse | ClickHouse | 654,094,530 | 12,160 | azat | Cc @alexey-milovidov | 2020-07-06T08:31:32 | 2020-07-06T08:31:32 | {} | MEMBER |
ClickHouse | ClickHouse | 654,183,393 | 12,153 | alesapin | Performance is not affected. | 2020-07-06T11:43:50 | 2020-07-06T11:43:50 | {} | MEMBER |
ClickHouse | ClickHouse | 654,258,923 | 9,899 | filimonov | Loosely related to #7666 | 2020-07-06T14:03:12 | 2020-07-06T14:03:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,260,363 | 7,772 | filimonov | STill true on current master (20.6.1) | 2020-07-06T14:05:37 | 2020-07-06T14:05:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,284,625 | 12,158 | alexey-milovidov | > Is this trace from the failed stress test check?
Yes.
I download server log and grep for `###`. | 2020-07-06T14:48:59 | 2020-07-06T14:48:59 | {} | MEMBER |
ClickHouse | ClickHouse | 654,294,228 | 4,007 | filimonov | In general such a message means that clickhouse was waiting for too long time in uninterruptible kernel mode.
> CAUSE:"Task blocked" messages indicate that a process has become blocked in uninterruptible kernel mode for longer than a specified timeout. The timeout is defined by kernel parameter hung_task_timeout_se... | 2020-07-06T15:05:05 | 2020-07-06T15:05:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,461,285 | 11,892 | Akazz | Performance tests failures seem unrelated completely.
fyi, @akuzm | 2020-07-06T20:57:14 | 2020-07-06T20:57:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,492,107 | 12,179 | azat | >I wish this was true, but it is required for dictGet* functions (since it needs attribute name to detect injectiveness)
@alexey-milovidov creating proper sample block here will be overkill, how about convert `isInjective(Block)` to `isInjetive(std::optional<Block>)` and done with it? | 2020-07-06T22:15:20 | 2020-07-06T22:15:28 | {} | MEMBER |
ClickHouse | ClickHouse | 654,128,287 | 12,142 | akuzm | https://twitter.com/trav_downs/status/1262058940831092738
A nice trick they use in another sampling memory profiler to record deallocation for the same blocks. | 2020-07-06T09:39:33 | 2020-07-06T09:39:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,227,909 | 8,071 | filimonov | One more symptom of the same bug reported:
```
2020.07.03 10:41:25.431186 [ 450 ] {} <Fatal> BaseDaemon: ########################################
2020.07.03 10:41:25.431247 [ 450 ] {} <Fatal> BaseDaemon: (version 19.15.3.6) (from thread 305) Received signal Segmentation fault (11).
2020.07.03 10:41:25.431263 [ 45... | 2020-07-06T13:12:56 | 2020-07-06T13:12:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,297,232 | 11,933 | den-crane | And it's actually strange 2**20/2907 = 360 bytes.
It means that a length of a part name is about 360 bytes.
@asurovenko can you show
`select partition, name from system.parts where active=1 and table='...' limit 10`
| 2020-07-06T15:10:29 | 2020-07-06T15:10:51 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 541,551,784 | 7,124 | choucmei | > Hard to follow.
>
> > and when i executed sql this,and all data has been deleted
> > `alter table test_01 delete where id in (select id from test_01)`
>
> I would also say it should delete everything :)
>
> Can you please provide some simple/synthetic example? We need simple table(s) definition, some data a... | 2019-10-14T08:28:34 | 2020-07-06T15:12:12 | {} | NONE |
ClickHouse | ClickHouse | 654,410,106 | 7,347 | filimonov | Related https://github.com/ClickHouse/ClickHouse/issues/9088#issuecomment-611062493 | 2020-07-06T18:58:13 | 2020-07-06T18:58:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,447,079 | 12,171 | alexey-milovidov | > Will it work w/o debug symbols installed?
Yes. | 2020-07-06T20:24:04 | 2020-07-06T20:24:04 | {} | MEMBER |
ClickHouse | ClickHouse | 654,455,277 | 12,178 | alexey-milovidov | No objections, let's upgrade.
And also the author of Arrow has said that it does not need flatbuffers anymore. | 2020-07-06T20:43:05 | 2020-07-06T20:43:05 | {} | MEMBER |
ClickHouse | ClickHouse | 654,471,285 | 7,347 | filimonov | Seems fixed in https://github.com/ClickHouse/ClickHouse/pull/11807 | 2020-07-06T21:20:12 | 2020-07-06T21:20:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 653,991,042 | 10,405 | alexey-milovidov | @CurtizJ Did you manage to reproduce with max_threads = 24? | 2020-07-06T03:06:19 | 2020-07-06T03:06:19 | {} | MEMBER |
ClickHouse | ClickHouse | 654,039,577 | 12,157 | alexey-milovidov | ```
milovidov@milovidov-desktop:~/work/ClickHouse/tests/queries/0_stateless$ for i in {1..1000}; do clickhouse-client -n < 00721_force_by_identical_result_after_merge_zookeeper.sql; done | grep -vP '0$'
milovidov@milovidov-desktop:~/work/ClickHouse/tests/queries/0_stateless$ for i in {1..10000}; do clickhouse-client ... | 2020-07-06T06:23:30 | 2020-07-06T06:23:30 | {} | MEMBER |
ClickHouse | ClickHouse | 654,255,401 | 12,127 | thyn | SummingMergeTree - Thanks. I've fixed it by changing engine to AggregatingMergeTree
example:
1. create table
```
CREATE TABLE temp.test4 (
d Date,
f1 SimpleAggregateFunction(anyLast, UInt8),
f2 SimpleAggregateFunction(anyLast, UInt8)
)
ENGINE = AggregatingMergeTree()
PARTITION BY tuple()
... | 2020-07-06T13:57:21 | 2020-07-06T14:26:50 | {} | NONE |
ClickHouse | ClickHouse | 654,273,350 | 5,847 | filimonov | Old ticket.
Most possible reason mentioned in that comment:
https://github.com/ClickHouse/ClickHouse/issues/5847#issuecomment-508250600
If if will be reproduced on newer versions - please feel free to open new one. | 2020-07-06T14:28:29 | 2020-07-06T14:28:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,288,304 | 8,927 | alexey-milovidov | It works as documented. Slightly more than `max_result_rows` can be returned if more rows are in the same block.
Don't mix up "query complexity settings" with the LIMIT clause.
If you want to limit the number of rows, you should use LIMIT clause.
When you use query complexity settings, the server will check thes... | 2020-07-06T14:55:07 | 2020-07-06T14:55:07 | {} | MEMBER |
ClickHouse | ClickHouse | 653,988,354 | 9,088 | alexey-milovidov | > SELECT arrayJoin([0, 1, 3, NULL]) AS x, x = 0, if(x = 0, 'x=0', 'x<>0');
@filimonov The query result is totally correct and it does not relate to the issue. | 2020-07-06T02:54:18 | 2020-07-06T02:54:18 | {} | MEMBER |
ClickHouse | ClickHouse | 654,066,549 | 12,041 | abyss7 | > I'm only worried in WARNING messages during configuration stage. We have them for more than a week.
> [#11300 (comment)](https://github.com/ClickHouse/ClickHouse/pull/11300#discussion_r446620461)
>
> As long as these messages are gone, I'm Ok with merging this PR.
>
> @abyss7 You can continue in another PR.
... | 2020-07-06T07:32:01 | 2020-07-06T07:32:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,296,321 | 6,879 | filimonov | That can be addressed (at least partially) with `os_thread_priority` | 2020-07-06T15:08:53 | 2020-07-06T15:09:23 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,302,654 | 1,946 | filimonov | Very old issue, no reproducible sample provided. A lot of changes happened in the meantime.
If you will face it on newer clickhouse versions - please feel free to open a new one. | 2020-07-06T15:20:25 | 2020-07-06T15:20:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,330,203 | 8,706 | den-crane | @ucasFL clickhouse-client --date_time_input_format=basic | 2020-07-06T16:10:44 | 2020-07-06T16:10:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 654,445,648 | 12,174 | alexey-milovidov | Builds Ok. | 2020-07-06T20:20:38 | 2020-07-06T20:20:38 | {} | MEMBER |
ClickHouse | ClickHouse | 654,504,007 | 12,142 | alexey-milovidov | @akuzm Now it is possible to track every allocation without patching ClickHouse, just look at
`max_untracked_memory`, `total_memory_profiler_step`... | 2020-07-06T22:55:47 | 2020-07-06T22:55:47 | {} | MEMBER |
ClickHouse | ClickHouse | 648,083,719 | 11,741 | 4ertus2 | Optimisation disabled in 20.5 branch and not present in 20.4. | 2020-06-23T11:23:37 | 2020-06-23T11:23:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 648,109,277 | 11,334 | nikitamikhaylov | > Or with small epsilon?
One naive approach is to add Laplace noise of scale 1/epsilon. And Laplace distribution with b as second parameter has variance equals to 2b^2. Other approaches have the same semantics.
Also I will leave [paper](https://arxiv.org/pdf/1909.01917.pdf) here. | 2020-06-23T12:20:15 | 2020-06-23T12:20:15 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.