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 | 605,857,077 | 9,923 | azat | >Functional stateless tests (release, streams) — fail: 38, passed: 74
>Functional stateless tests (thread) — fail: 20, passed: 344, skipped: 5
No errors in clickhouse-server.log, OOM/reboot?
(Other tests fails in upstream/master too and performance looks unrelated) | 2020-03-30T08:27:28 | 2020-03-30T08:27:28 | {} | MEMBER |
ClickHouse | ClickHouse | 606,201,190 | 9,927 | SaltTan | I don't know why, but here it is:
Speedup ddl cleanup #4502 | 2020-03-30T19:33:36 | 2020-03-30T19:33:36 | {
"confused": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,324,192 | 9,836 | hustnn | @den-crane
Yes, you are right. All needed parts will be provided by the coordinator to the CH instance. That is why I am considering to make the query execution in the CH instance pipelined with the transfer of these parts. | 2020-03-31T00:29:45 | 2020-03-31T00:29:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,751,012 | 9,923 | azat | And seems that performance tests is unstable too (since it shows `Performance 3 faster, 35 unstable, 5 errors` in upstream/master), and plus this PR should not affect them, since it does not changes existing code paths. | 2020-03-31T16:58:58 | 2020-03-31T16:58:58 | {} | MEMBER |
ClickHouse | ClickHouse | 606,794,439 | 9,961 | alexey-milovidov | And also the question: why TestKeeper session can be expired? | 2020-03-31T18:24:35 | 2020-03-31T18:24:35 | {} | MEMBER |
ClickHouse | ClickHouse | 605,608,925 | 9,113 | azat | What is the status of this one? Does it postponed? (I have some stuff that will require this) | 2020-03-29T09:28:40 | 2020-03-29T09:28:40 | {} | MEMBER |
ClickHouse | ClickHouse | 605,763,305 | 9,836 | den-crane | To execute any query CH needs knowledge about all parts of the tables this query touches.
I do not see any flaws in current architecture. What is your problem? | 2020-03-30T03:29:49 | 2020-03-30T14:39:51 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,042,920 | 9,902 | alexey-milovidov | You can rewrite the `flatc` binary in `CMakeFiles` with your own binary and continue build. | 2020-03-30T14:43:37 | 2020-03-30T14:43:37 | {} | MEMBER |
ClickHouse | ClickHouse | 606,709,651 | 9,969 | genzgd | We've noticed this issue only happens on large payloads. The Brotli response crashes at about 11MB, while the same gzip response completes successfully at 57MB. if we ask for a smaller JSON dataset the Brotli request completes successfully. | 2020-03-31T15:47:30 | 2020-03-31T21:02:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,694,746 | 9,923 | alexey-milovidov | We also should pay attention to two level GROUP BY and MergingAggregatedMemoryEfficient.
In this mode, every block has bucket_num assigned and data is merged per buckets. | 2020-03-29T20:20:40 | 2020-03-29T20:20:40 | {} | MEMBER |
ClickHouse | ClickHouse | 605,695,586 | 9,919 | azat | >readability-identifier-naming
Interesting, why it does not report the same for totalBytes and totalRows | 2020-03-29T20:25:59 | 2020-03-29T20:25:59 | {} | MEMBER |
ClickHouse | ClickHouse | 605,867,560 | 9,923 | azat | >but GLOBAL IN does not works even now (i.e. distributed_group_by_no_merge), fails with LOGACAL_ERROR Table expression is undefined, Method: ExpressionAnalyzer::interpretSubquery., which then will abort in debug build
Added a test into `bugs` | 2020-03-30T08:49:05 | 2020-03-30T08:49:05 | {} | MEMBER |
ClickHouse | ClickHouse | 605,941,023 | 9,928 | filimonov | You can configure executable dictionaries only via XML configuration file.
Otherwise, every DB user will have a possibility to execute some malicious code on your ClickHouse server.
See https://github.com/ClickHouse/ClickHouse/pull/8127/ | 2020-03-30T11:20:35 | 2020-03-30T11:20:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,950,919 | 6,185 | alexey-milovidov | This task is worth doing. @KochetovNicolai is currently trying to beat the hopelessness. | 2020-03-30T11:44:17 | 2020-03-30T11:44:17 | {} | MEMBER |
ClickHouse | ClickHouse | 606,514,488 | 9,946 | alex-zaitsev | The same problem is reproduced with TTL deletes -- if TTL drops a partition with data not in quorum, quorum gets stuck forever. | 2020-03-31T09:41:58 | 2020-03-31T09:58:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,760,974 | 5,997 | den-crane | https://t.me/clickhouse_ru/150955
_Ntfs, докер, винда. Уже разобрался. Вольюм с данными БД коряво прокинул. Самое интересное в том что порой проходило, порой нет)
clickhouse-server:
image: yandex/clickhouse-server
restart: always
ulimits:
nofile:
soft: 262144
h... | 2020-03-31T17:17:26 | 2020-03-31T17:17:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,674,636 | 9,919 | azat | >Integration tests (asan) — failed:1, passed:0, error:0
`==================== 2 passed, 2 warnings in 76.88 seconds =====================`
Hm, interesting. | 2020-03-29T17:54:25 | 2020-03-29T17:54:25 | {} | MEMBER |
ClickHouse | ClickHouse | 605,678,578 | 9,923 | alexey-milovidov | We can also test queries with GLOBAL IN/JOIN but don't worry - even if they don't work in Distributed over Distributed, it's still acceptable. | 2020-03-29T18:24:13 | 2020-03-29T18:24:13 | {} | MEMBER |
ClickHouse | ClickHouse | 605,949,560 | 6,286 | alexey-milovidov | @stavrolia refused to finish this task. | 2020-03-30T11:41:04 | 2020-03-30T11:41:04 | {
"confused": 4
} | MEMBER |
ClickHouse | ClickHouse | 606,899,527 | 9,957 | filimonov | Materialized views work like AFTER INSERT trigger, copying block which arrived to destination table.
Usually MV is build using Summing- / Aggregating- / Collapsing- MergeTree which will do some extra processing of duplicates.
In case you need the VIEW to rebuild on every select time - maybe usual VIEW will work bet... | 2020-03-31T21:59:41 | 2020-03-31T21:59:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,985,044 | 9,955 | alexey-milovidov | I prefer more simple solution - to acquire DDL guards one by one in SYSTEM RESTART REPLICAS and allow failure in the middle. | 2020-04-01T02:10:26 | 2020-04-01T02:10:26 | {} | MEMBER |
ClickHouse | ClickHouse | 605,609,260 | 9,113 | alexey-milovidov | The task is active, worst case estimate is May 2020, but @dimarub2000 should target for April. | 2020-03-29T09:32:08 | 2020-03-29T09:32:08 | {
"+1": 3
} | MEMBER |
ClickHouse | ClickHouse | 605,644,297 | 9,918 | den-crane | 1. tcp has slightly better performance
2. yes | 2020-03-29T14:24:04 | 2020-03-29T14:24:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,684,785 | 9,923 | azat | >We can also test queries with GLOBAL IN/JOIN
JOIN works, but GLOBAL IN does not works even now (i.e. `distributed_group_by_no_merge`), fails with LOGACAL_ERROR `Table expression is undefined, Method: ExpressionAnalyzer::interpretSubquery.`, which then will abort in debug build | 2020-03-29T19:11:25 | 2020-03-29T19:12:15 | {} | MEMBER |
ClickHouse | ClickHouse | 605,685,715 | 9,923 | azat | Oh, I see, looks like the failed tests are failed on upstream/master too | 2020-03-29T19:18:57 | 2020-03-29T19:18:57 | {} | MEMBER |
ClickHouse | ClickHouse | 605,693,191 | 9,499 | 4ertus2 | You also need to implement `*Overflow()` functions for 256-bit types here https://github.com/ClickHouse/ClickHouse/blob/master/base/common/arithmeticOverflow.h#L30
| 2020-03-29T20:11:04 | 2020-03-29T20:11:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,698,527 | 9,923 | azat | >We also should pay attention to two level GROUP BY and MergingAggregatedMemoryEfficient.
In this mode, every block has bucket_num assigned and data is merged per buckets.
Looks ok (test updated) | 2020-03-29T20:44:29 | 2020-03-29T20:44:29 | {} | MEMBER |
ClickHouse | ClickHouse | 605,852,984 | 9,919 | azat | So the following tests fails now, and AFAICS all of them are either flacky or fails in upstream/master too:
- Integration tests -- fails in upstream/master too
- `Functional stateless tests (ubsan) — fail: 4` -- fails in upstream/master too (all 4 matched)
- `01098_temporary_and_external_tables` -- looks "flacky", s... | 2020-03-30T08:18:55 | 2020-03-30T08:23:32 | {} | MEMBER |
ClickHouse | ClickHouse | 605,975,368 | 9,908 | filimonov | It looks like it also resolves #9637 and makes that #9455 unneeded .
Is that correct? | 2020-03-30T12:41:39 | 2020-03-30T12:41:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,491,242 | 9,902 | murazh | > You can rewrite the `flatc` binary in `CMakeFiles` with your own binary and continue build.
I have replaced all files in /ClickHouse/build/contrib/flatbuffers with compiled files from /flatbuffers/CMakeFile and also added flatc binary but error the same | 2020-03-31T08:55:18 | 2020-03-31T08:55:18 | {} | NONE |
ClickHouse | ClickHouse | 606,894,415 | 9,969 | filimonov | Are you able to reproduce that using some simple synthetic example? | 2020-03-31T21:47:11 | 2020-03-31T21:47:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,979,216 | 9,967 | alexey-milovidov | After we changed docker image for build, It's also required to port the fix for "Compatibility check". | 2020-04-01T01:51:49 | 2020-04-01T01:51:49 | {} | MEMBER |
ClickHouse | ClickHouse | 606,979,432 | 9,967 | alexey-milovidov | https://github.com/ClickHouse/ClickHouse/pull/9833 | 2020-04-01T01:52:29 | 2020-04-01T01:52:29 | {} | MEMBER |
ClickHouse | ClickHouse | 605,749,012 | 9,836 | hustnn | @den-crane Understand.
That is one example, as you mentioned, as the part loading consumes a lot of time, if the query execution can be pipelined with the loading phase, the end-to-end time may be a little longer than the loading time. Otherwise, the end-to-end will be the loading time + execution time.
How do ... | 2020-03-30T02:27:25 | 2020-03-30T02:27:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,654,344 | 9,919 | alexey-milovidov | It's fine :) | 2020-03-29T15:29:20 | 2020-03-29T15:29:20 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 605,656,134 | 9,919 | azat | Ok, PR description updated (to reflect this info at least). | 2020-03-29T15:40:53 | 2020-03-29T15:40:53 | {} | MEMBER |
ClickHouse | ClickHouse | 605,946,028 | 9,789 | alexey-milovidov | This part has failed. | 2020-03-30T11:32:42 | 2020-03-30T11:32:42 | {} | MEMBER |
ClickHouse | ClickHouse | 605,948,279 | 8,561 | alexey-milovidov | @KochetovNicolai did not review in time.
The optimization is great but has limited use - can be postponed. | 2020-03-30T11:37:55 | 2020-03-30T11:38:01 | {} | MEMBER |
ClickHouse | ClickHouse | 606,044,360 | 9,908 | alexey-milovidov | Yes. And also it reverts https://github.com/ClickHouse/ClickHouse/pull/1186 | 2020-03-30T14:46:06 | 2020-03-30T14:46:06 | {
"heart": 1
} | MEMBER |
ClickHouse | ClickHouse | 606,057,089 | 9,870 | genzgd | Yes, we definitely had some weirdness because we upgrade to 19.17 at one point and then downgraded back to 19.3.3, so there was some unexpected inconsistency in zookeeper metadata definitions around index granularity. Thanks for pointing us in the right direction! | 2020-03-30T15:07:32 | 2020-03-30T15:07:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,578,255 | 9,961 | alesapin | Previously, if TestKeeper session expires for some reason we will not be able to reconnect because `startNewSessions` doesn't pass implementation name to the constructor and use the default `"zookeeper". | 2020-03-31T11:48:27 | 2020-03-31T11:48:27 | {} | MEMBER |
ClickHouse | ClickHouse | 606,904,222 | 9,410 | excitoon | Rebased it, fixed conflicts and the comment. | 2020-03-31T22:11:52 | 2020-03-31T22:11:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,915,037 | 9,940 | excitoon | I guess there are two issues here:
1) Corrupted part data upon deletion (missing `checksum.txt`);
2) `Sign` was somehow collapsed to `-37`, which is not good at all. | 2020-03-31T22:24:48 | 2020-03-31T22:24:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,480,311 | 9,912 | blinkov | @FedericoCeratto btw you could just browse through documentation sources, they are more or less readable even on GitHub: https://github.com/ClickHouse/ClickHouse/tree/v18.16/docs/en | 2020-03-28T16:27:31 | 2020-03-28T16:27:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,533,842 | 9,915 | knoguchi | ouch. I overlooked it. Thanks! | 2020-03-28T23:36:16 | 2020-03-28T23:36:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,694,520 | 9,923 | alexey-milovidov | > B will do nothing except forwarding blocks.
As I understand, the main challenge is that it cannot just forward blocks from different servers as a single stream... because it will interfere with merge sorting. Do I understand correctly that this is the reason why you are doing additional intermediate sorting? | 2020-03-29T20:19:16 | 2020-03-29T20:19:38 | {} | MEMBER |
ClickHouse | ClickHouse | 605,786,204 | 9,836 | hustnn | @den-crane Thanks for your reply.
Yes, currently, CH needs knowledge about all parts of the tables this query touches.
I did some changes on ClickHouse, now the parts is distributed through a so-called coordinator, the distribution may take some time. If still need all parts ready, the query can not start until a... | 2020-03-30T05:12:53 | 2020-03-30T05:12:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,990,351 | 9,902 | murazh | > It means that your system does not support CPU instruction set that we use for ClickHouse (SSE 4.2).
> It happens on build time because the intermediate build tool `flatc` is also build with this instruction set but it cannot run.
>
> This is build issue. Intermediate tools must be built with the CPU instruction ... | 2020-03-30T13:12:44 | 2020-03-30T13:12:44 | {} | NONE |
ClickHouse | ClickHouse | 606,032,902 | 9,928 | zcw5116 | > You can configure executable dictionaries only via XML configuration file.
> Otherwise, every DB user will have a possibility to execute some malicious code on your ClickHouse server.
> See #8127
Thank you very much | 2020-03-30T14:26:53 | 2020-03-30T14:26:53 | {} | NONE |
ClickHouse | ClickHouse | 606,084,080 | 9,945 | den-crane | 1. memory leak https://github.com/ClickHouse/ClickHouse/issues/3976
set in users profile compile_expressions=0
2. reduce config.xml: <mark_cache_size>5368709120</mark_cache_size>
from 5GB to 500MB
3. Add swap 16 GB | 2020-03-30T15:53:01 | 2020-03-30T15:53:51 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,364,381 | 9,932 | testlujing | > That looks like one more reincarnation of the old bug (lack of query context).
> It should be easy to fix, in the meanwhile you will need to use some workaround (like do the join before).
my clickhouse version is ClickHouse 19.17.4.11
how to do the jion before? using MATERIALIZED VIEW?
but do not update previou... | 2020-03-31T02:40:53 | 2020-03-31T03:06:40 | {} | NONE |
ClickHouse | ClickHouse | 606,806,745 | 7,247 | den-crane | Говорят непонятно как теперь апдейты делать.
> Здравствуйте! Вы мне както подсказывали как в кликхаусе делаете мутации с условиями через словари:
>
> Denny Crane [not a Yandex bot], [22.01.20 04:27]
> такие запросы нельзя делать, alter для каждой строки xxx проверить where.
>
> Я готовлю таблицу ReplicatedMer... | 2020-03-31T18:48:55 | 2020-03-31T18:50:24 | {
"confused": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,819,431 | 9,951 | qoega | Commits to documentation do not fire CI builds for binaries etc. Thus makes them greener than master itself. | 2020-03-31T19:13:33 | 2020-03-31T19:13:33 | {} | MEMBER |
ClickHouse | ClickHouse | 606,838,386 | 9,927 | alexey-milovidov | Ok, let's leave separate thread. | 2020-03-31T19:50:11 | 2020-03-31T19:50:11 | {} | MEMBER |
ClickHouse | ClickHouse | 606,930,083 | 9,841 | ghost | It is merged at https://github.com/ClickHouse/ClickHouse/pull/9938. | 2020-03-31T23:03:40 | 2020-03-31T23:03:40 | {} | NONE |
ClickHouse | ClickHouse | 606,940,674 | 9,969 | genzgd | This is the table definition:
```CREATE TABLE analytics.kpipe_metrics (`datetime` DateTime, `pipeline` String, `host` String, `metric` String, `eps` UInt64, `cnt` UInt64) ENGINE = ReplicatedMergeTree('/clickhouse/tables/{shard}/analytics.kpipe_metrics', '{replica}') PARTITION BY toDate(datetime) ORDER BY (datetime,... | 2020-03-31T23:33:16 | 2020-03-31T23:33:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,500,740 | 9,854 | filimonov | > Fixed automatically in 20.1:
#8311, 272b7b4
According to reporter the problem reproduces on 20.1 too.
> Reproduced several times on ClickHouse version 19.16.10.44 and later on 20.1.6.30
| 2020-03-28T18:31:20 | 2020-03-28T18:32:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,533,715 | 9,915 | den-crane | Column names must be the same in a MV select and the target (TO) table
CREATE MATERIALIZED VIEW test.a2b TO test.b AS
SELECT
toStartOfHour(t) **as t**
FROM test.a;
MV uses column names when it does insert into the target table. | 2020-03-28T23:34:40 | 2020-03-28T23:36:03 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,694,967 | 9,919 | alexey-milovidov | Minor:
```
2020-03-29 17:18:04 /build/obj-x86_64-linux-gnu/../dbms/src/Storages/StorageBuffer.cpp:730:27: error: invalid case style for local variable 'underlyingRows' [readability-identifier-naming,-warnings-as-errors]
2020-03-29 17:18:04 std::optional<UInt64> underlyingRows;
2020-03-29 17:18:04 ... | 2020-03-29T20:21:56 | 2020-03-29T20:21:56 | {} | MEMBER |
ClickHouse | ClickHouse | 605,838,933 | 8,979 | gontarzpawel | @alesapin Hello, did you get some time to take a look at this issue? | 2020-03-30T07:48:08 | 2020-03-30T07:48:08 | {} | NONE |
ClickHouse | ClickHouse | 606,059,976 | 616 | abfisher0417 | This worked like a charm. I will also add that `clickhouse-server` needs to be restarted after manually renaming directories to include the password; otherwise, you'll see something like: `Distributed.DirectoryMonitor: Poco::Exception. Code: 1000 File not found` in the logs. | 2020-03-30T15:12:38 | 2020-03-30T15:12:38 | {} | NONE |
ClickHouse | ClickHouse | 606,159,558 | 9,927 | garneja | I don't know why DDLWorker and DDLWorkerClnr exist as two distinct threads. There's no concurrency benefit here, that I can see. We saw another crash of the same nature when the zookeeper session expires. It would we wise to combine the two DDLWorker and DDLWorkerClnr into one thread. I'll post changes to combine the t... | 2020-03-30T18:16:32 | 2020-03-30T18:16:32 | {} | NONE |
ClickHouse | ClickHouse | 606,571,766 | 9,864 | 4ertus2 | Related #9910 | 2020-03-31T11:33:15 | 2020-03-31T11:33:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,980,179 | 9,967 | alexey-milovidov | Also https://github.com/ClickHouse/ClickHouse/pull/9433 can be relevant. | 2020-04-01T01:54:39 | 2020-04-01T01:54:39 | {} | MEMBER |
ClickHouse | ClickHouse | 606,599,553 | 9,964 | blinkov | nowadays there's also es+fr docs folders | 2020-03-31T12:36:08 | 2020-03-31T12:36:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,928,625 | 9,969 | genzgd | I've attempted to reproduce this by returning a similar amount of data querying the system.numbers table with some constant columns. That did not work -- a Brotli compressed response of 50+ megabytes worked fine. I can reproduce it will similar exception results against the real table, with the exception thrown at ap... | 2020-03-31T22:59:37 | 2020-03-31T22:59:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,622,147 | 9,905 | CurtizJ | Previously system tables were written in wide format and their column sizes were counted. Also only TSan check was affected, because it skips the only two tests, in which tables with adaptive granularity are created. | 2020-03-29T11:32:09 | 2020-03-29T11:32:09 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 605,938,668 | 9,932 | filimonov | That looks like one more reincarnation of the old bug (lack of query context).
It should be easy to fix, in the meanwhile you will need to use some workaround (like do the join before). | 2020-03-30T11:15:04 | 2020-03-30T11:15:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,015,208 | 154 | alexey-milovidov | This task is easy to do but there is very low demand for this feature. | 2020-03-30T13:56:54 | 2020-03-30T13:56:54 | {} | MEMBER |
ClickHouse | ClickHouse | 606,168,978 | 9,544 | shayrybak | I can confirm those settings work much better (similar to older version).
I'm going to close the case, but maybe the defaults should change to reflect that? | 2020-03-30T18:34:33 | 2020-03-30T18:34:33 | {} | NONE |
ClickHouse | ClickHouse | 606,448,833 | 9,923 | azat | Test failures looks unrelated.
> test_settings_constraints
> 01098_temporary_and_external_tables | 2020-03-31T07:23:29 | 2020-03-31T07:23:29 | {} | MEMBER |
ClickHouse | ClickHouse | 606,720,386 | 4,810 | siradjev | > We don't have wrong query result in newer versions.
But we still lack of type conversions for IN subquery.
Seems it's also not working for pruning. Here is example of DateTime and Date, when partitioning key is DateTime column, but actual condition on partitioning key is Date type.
```
drop table if exists t... | 2020-03-31T16:01:49 | 2020-03-31T16:02:44 | {} | NONE |
ClickHouse | ClickHouse | 606,790,016 | 9,666 | oandrew | @alexey-milovidov
https://github.com/ClickHouse-Extras/poco/pull/19
Do we want to have such changes in the fork or upstream first ( i.e.
https://github.com/pocoproject/poco) ? | 2020-03-31T18:15:20 | 2020-03-31T18:17:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,981,482 | 9,962 | alexey-milovidov | Don't understant why it became broken. | 2020-04-01T01:58:24 | 2020-04-01T01:58:24 | {} | MEMBER |
ClickHouse | ClickHouse | 605,531,170 | 9,912 | FedericoCeratto | Thanks! | 2020-03-28T23:02:53 | 2020-03-28T23:02:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,651,857 | 9,919 | azat | > 00341_squashing_insert_select2
This is due to incorrect detection of `optimize_trivial_count_query`, fixed too (can be moved into separate PR, since it is completely unrelated to this one, but it's up to you) | 2020-03-29T15:12:41 | 2020-03-29T15:12:41 | {} | MEMBER |
ClickHouse | ClickHouse | 605,678,390 | 9,923 | alexey-milovidov | Do I understand correctly that it forwards data when WithMergeableState -> WithMergeableState is requested and possibly, applies additional pre-sorting/distrinct/limit (but for what purpose?) | 2020-03-29T18:22:41 | 2020-03-29T18:22:41 | {} | MEMBER |
ClickHouse | ClickHouse | 605,694,025 | 9,923 | alexey-milovidov | > AFAIU It does not pre-sorting it sorts results that had been received so far from the shards, since otherwise this step will be skipped and only the final WithMergeableState -> Complete will sort, and not complete sort but only merge sorted already results.
If we have three servers:
A (local) -> B (distributed,... | 2020-03-29T20:15:53 | 2020-03-29T20:15:53 | {} | MEMBER |
ClickHouse | ClickHouse | 605,694,197 | 9,923 | alexey-milovidov | PS. We can do re-merging intermediate data on intermediate servers but it's out of scope of this task. | 2020-03-29T20:17:10 | 2020-03-29T20:17:10 | {} | MEMBER |
ClickHouse | ClickHouse | 605,945,397 | 9,940 | qoega | @excitoon | 2020-03-30T11:31:06 | 2020-03-30T11:31:06 | {} | MEMBER |
ClickHouse | ClickHouse | 606,523,619 | 5,997 | alanlans | > > I have same problem
>
> What type of filesystem do you use?
NTFS
2020-03-31 11:39:58] Connected
default> CREATE TABLE mytable(
some_date Date,
some_int UInt64,
some_str String,
some_float Float32
) ENGINE = MergeTree(some_date, (some_int, some_date), 8... | 2020-03-31T09:59:18 | 2020-03-31T09:59:18 | {} | NONE |
ClickHouse | ClickHouse | 606,617,253 | 9,944 | vzakaznikov | @alexey-milovidov, I think this PR can be merged. The failing tests have nothing to do with this small test change. The test itself is passing. | 2020-03-31T13:11:27 | 2020-03-31T13:11:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,819,070 | 9,961 | alesapin | > And also the question: why TestKeeper session can be expired?
I have connected to sever with GDB. After I have detached from the server all sessions were expired and were not able to recover. | 2020-03-31T19:12:48 | 2020-03-31T19:19:56 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 606,836,394 | 9,941 | alexey-milovidov | > wait for tests and merge
Test did not finish within a day: `Docs check Pending — Started`
> approved these changes yesterday
There is no explanation for the motivation of this function. What is the "top level continent"?
This is a classical example of documentation that only makes things worse. | 2020-03-31T19:46:38 | 2020-03-31T19:47:09 | {} | MEMBER |
ClickHouse | ClickHouse | 606,895,895 | 9,960 | filimonov | Please update ClickHouse for example to v20.1.9.54-stable
See https://github.com/ClickHouse/ClickHouse/issues/9153#issuecomment-586965006 | 2020-03-31T21:50:43 | 2020-03-31T21:51:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,682,038 | 9,923 | azat | >Do I understand correctly that it forwards data when WithMergeableState -> WithMergeableState is requested and possibly, applies additional pre-sorting/distrinct/limit (but for what purpose?)
AFAIU It does not pre-sorting it sorts results that had been received so far from the shards, since otherwise this step will... | 2020-03-29T18:49:47 | 2020-03-29T18:49:47 | {} | MEMBER |
ClickHouse | ClickHouse | 605,933,281 | 9,923 | alexey-milovidov | Ok. But it also means that memory-efficient aggregation won't work without remerging.
You can test the case when servers send data by buckets... | 2020-03-30T11:02:13 | 2020-03-30T11:02:13 | {} | MEMBER |
ClickHouse | ClickHouse | 606,000,497 | 7,278 | alexey-milovidov | The code looks Ok. I will try to merge... | 2020-03-30T13:31:18 | 2020-03-30T13:31:18 | {} | MEMBER |
ClickHouse | ClickHouse | 606,034,318 | 9,836 | den-crane | Then your coordinator should have / store information about all parts ( parts metadata ) and be able to provide this metadata to CH instances. | 2020-03-30T14:29:20 | 2020-03-30T14:29:20 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,168,644 | 9,927 | alexey-milovidov | I agree, let's do that! | 2020-03-30T18:33:53 | 2020-03-30T18:33:53 | {} | MEMBER |
ClickHouse | ClickHouse | 606,971,277 | 9,957 | testlujing | > Materialized views work like AFTER INSERT trigger, copying block which arrived to destination table.
> Usually MV is build using Summing- / Aggregating- / Collapsing- MergeTree which will do some extra processing of duplicates.
>
> In case you need the VIEW to rebuild on every select time - maybe usual VIEW will ... | 2020-04-01T01:21:53 | 2020-04-01T01:21:53 | {} | NONE |
ClickHouse | ClickHouse | 605,539,250 | 9,905 | CurtizJ | > Ok, but what's wrong with stateful test?
> There should be non-zero values in system.columns because test.hits table exists.
See #9916
| 2020-03-29T00:40:36 | 2020-03-29T00:40:36 | {} | MEMBER |
ClickHouse | ClickHouse | 605,544,243 | 9,905 | alexey-milovidov | Ok... but how it is related to changing the threshold of wide parts of system tables? | 2020-03-29T01:40:46 | 2020-03-29T01:40:46 | {} | MEMBER |
ClickHouse | ClickHouse | 605,700,455 | 9,923 | azat | > As I understand, the main challenge is that it cannot just concatenate blocks from different servers to a single stream... because it will interfere with merge sorting. Do I understand correctly that this is the reason why you are doing additional intermediate sorting?
Yes, the problem is that intermediate server ... | 2020-03-29T20:58:49 | 2020-03-29T20:58:49 | {} | MEMBER |
ClickHouse | ClickHouse | 605,927,798 | 9,623 | akuzm | Just noticed this. We've discussed this offline with @qoega and @4ertus2 like three weeks ago. I don't remember much already, but the assertion failure means that the bloom filter creation code reinterpret_casts the field value from float to int, and god knows how it ever worked. I'm not going to fix it anytime soon, b... | 2020-03-30T10:49:59 | 2020-03-30T10:49:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,919,757 | 9,940 | excitoon | @qoega Which are `CREATE` statements and are there any ways to reproduce that? | 2020-03-31T22:32:43 | 2020-03-31T22:32:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 605,984,605 | 9,714 | 4ertus2 | Reproduced. Wrong result from remote shard:
```
SELECT 'distributed_1', _shard_num, toUInt8(1) AS dummy FROM d_one AS o WHERE o.dummy = 1 ORDER BY _shard_num;
```
```
distributed_1 2 1
```
The trouble is in QueryNormalizer that rewrites `o.dummy = 0` with ```(dummy AS `o.dummy`) = 0``` in `WHERE`
```
... | 2020-03-30T13:01:26 | 2020-03-30T13:21:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 606,179,709 | 9,923 | azat | >memory-efficient aggregation won't work without remerging.
Nice catch! Fixed and covered! | 2020-03-30T18:56:07 | 2020-03-30T18:56:07 | {} | MEMBER |
ClickHouse | ClickHouse | 606,539,128 | 146 | alesapin | Feature in this PR: https://github.com/ClickHouse/ClickHouse/pull/9948. | 2020-03-31T10:30:01 | 2020-03-31T10:30:33 | {} | MEMBER |
ClickHouse | ClickHouse | 602,705,687 | 9,829 | alexey-milovidov | > The CI errors seem irrelevant.
Yes, it's today feature.
Let's wait performance test... | 2020-03-23T16:20:27 | 2020-03-23T16:20:27 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.