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 | 531,175,309 | 1,768 | filimonov | Let's continue tracking that issue in #1562 | 2019-09-13T09:49:48 | 2019-09-13T09:49:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,298,379 | 5,843 | nvartolomei | @filimonov can't agree with none of the above. Especially considering that one accidental insert wrecks the table completely (until you alter the table). | 2019-09-13T16:10:06 | 2019-09-13T16:10:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,343,816 | 6,937 | akuzm | I'd prefer to call the Cell constructor in emplaceIfZero, because it would be only initialized on demand, and also would properly initialize complex cells like ClearableHashTableCell, which is not initialized to all zeros but has a version field. | 2019-09-13T18:29:00 | 2019-09-13T18:29:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,821,916 | 6,922 | den-crane | BTW, `Ternary operator always calculate two branches` -- not a bug but a performance optimization.
So you need to manage it somehow. | 2019-09-12T13:19:35 | 2019-09-12T13:19:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,837,440 | 3,490 | akuzm | I've tested this on 'AMD EPYC 7452 32-Core Processor'.
`SELECT 1 FROM test.perf_test GROUP BY toUInt16(intHash64(rand() % 10))` executes in ~35 ms on both 19.13 and 19.14. This is much faster that ~1s mentioned above, so I guess I'm not seeing the slowdown.
| 2019-09-12T13:54:59 | 2019-09-12T13:54:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,146,624 | 6,930 | alexey-milovidov | @Akazz to check "some queries hung" in functional tests (unrelated). | 2019-09-13T08:20:53 | 2019-09-13T08:20:53 | {} | MEMBER |
ClickHouse | ClickHouse | 531,157,462 | 6,766 | alexey-milovidov | @filimonov That's exactly how I found this issue. | 2019-09-13T08:53:47 | 2019-09-13T08:53:47 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 531,464,467 | 6,648 | ucasfl | Is the two arguments of this function fixed as a string and an integer? Or it can be a Column? | 2019-09-14T09:20:40 | 2019-09-14T09:20:40 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,708,672 | 4,299 | jasper-zhang | I met the same issue, and I found the added column can not be found in zk. | 2019-09-16T09:45:09 | 2019-09-16T09:45:09 | {} | NONE |
ClickHouse | ClickHouse | 531,744,455 | 6,937 | akuzm | > Is it some flaky test?
The logs are truncated, looks like some CI failure. | 2019-09-16T11:51:06 | 2019-09-16T11:51:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,774,280 | 6,947 | amosbird | I've heard we need `-gsplit-dwarf` or `-ggnu-pubnames ` to make `-Wl,-gdb-index` actually work | 2019-09-16T13:23:05 | 2019-09-16T13:23:05 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 530,821,171 | 6,922 | pachico | Thanks, I failed at finding duplicate issues. | 2019-09-12T13:17:40 | 2019-09-12T13:17:40 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,128,058 | 6,931 | jeffw-wherethebitsroam | I just did a docker pull and updated to ClickHouse server version 19.14.3 revision 54425. It seems to work as expected again on that version. Thanks. | 2019-09-13T07:18:58 | 2019-09-13T07:18:58 | {} | NONE |
ClickHouse | ClickHouse | 531,174,870 | 2,385 | filimonov | Let's continue tracking that issue in #1562 | 2019-09-13T09:48:21 | 2019-09-13T09:48:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,179,502 | 6,917 | filimonov | Similar https://github.com/yandex/ClickHouse/issues/6104 | 2019-09-13T10:03:45 | 2019-09-13T10:03:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,529,787 | 6,880 | andyyzh | @yingfeng
The test at that time was slow to use croaring serialization/deserialization,
and later, change to use clickhouse's own serialization/deserialization. | 2019-09-15T02:52:25 | 2019-09-15T02:52:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,535,371 | 6,880 | alexey-milovidov | @andyyzh @yingfeng You should add a performance test, so we can control it in CI. | 2019-09-15T05:13:14 | 2019-09-15T05:13:14 | {} | MEMBER |
ClickHouse | ClickHouse | 531,676,345 | 6,943 | filimonov | Hundreds of columns are ok.
Starting from let's say 1.5-2K columns performance degradation is expected, but rather should still work.
I've tried you case, and the problem reproduces easily.
```
zcat schema_20K_columns.sql.gz | clickhouse-client --max_query_size=1000000
laptop5591.local :) insert into schema_2... | 2019-09-16T07:57:35 | 2019-09-16T08:48:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,694,537 | 6,908 | nicelulu | The older versions of the table column with bitmap will not work properly. | 2019-09-16T08:59:55 | 2019-09-16T10:29:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,851,874 | 5,417 | akuzm | Remember the mysterious ubsan failure?
Good news: I can reproduce it every time under MemorySanitizer.
Bad news: It's in master now.
The old ubsan stack was like that, the new msan one is the same.
```
../dbms/src/Columns/ColumnArray.cpp:163:16: runtime error: pointer index expression with base 0xffffffffffffff... | 2019-09-12T14:28:04 | 2019-09-12T14:28:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,942,366 | 3,490 | akuzm | ```
$ uname -a
Linux amd-lab 4.19.62-13 #1 SMP Mon Jul 29 17:22:59 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
```
The distro is Ubuntu 19.04 | 2019-09-12T18:12:25 | 2019-09-12T18:13:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,150,818 | 6,766 | filimonov | That is very important for backward compatibility after #6265.
```
select * from system.query_log where event_date>=yesterday() and event_time > now() - 160 and query_id='check2' and type in (2,4)
Received exception from server (version 19.15.1):
Code: 53. DB::Exception: Received from localhost:9000. DB::Excep... | 2019-09-13T08:33:55 | 2019-09-13T08:33:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,167,016 | 5,017 | filimonov | Since 19.14 and #5273 :
```
CREATE TABLE test
(
`a` Int32,
`b` Enum8('b1' = 1, 'b2' = 2),
CONSTRAINT b_constraint CHECK b IN ('b1', 'b2')
)
ENGINE = MergeTree()
PARTITION BY tuple()
ORDER BY tuple()
Ok.
INSERT INTO test (a) values (1);
INSERT INTO test (a) VALUES
Received exception f... | 2019-09-13T09:22:59 | 2019-09-13T09:22:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,242,916 | 6,673 | abyss7 | Does it work on recent master? | 2019-09-13T13:45:15 | 2019-09-13T13:45:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,301,834 | 5,843 | alexey-milovidov | @filimonov Checking Enum values is very cheap comparing to INSERT. We should do it. | 2019-09-13T16:20:21 | 2019-09-13T16:20:21 | {} | MEMBER |
ClickHouse | ClickHouse | 530,794,243 | 6,807 | ildus | @proller this problem still exists on my system (latest arch linux), I'm not sure why you hardcoded `set (LLVM_PATHS "/usr/local/lib/llvm")` in cmake.
If I just add `find_package(LLVM REQUIRED CONFIG) ` (without paths) cmake can find LLVM without any problems.
My version of cmake is 3.15 | 2019-09-12T12:04:59 | 2019-09-12T12:05:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,813,467 | 6,912 | akuzm | > @akuzm thank you, going to investigate the issue. Also could you tell why style checker doesn't like these two lines?
>
> https://clickhouse-test-reports.s3.yandex.net/6912/20db1a75c6a241f6c55f56db9c337485022450fd/style_check.html
It's the tab character in the declaration, we only use spaces. | 2019-09-12T12:58:58 | 2019-09-12T12:58:58 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,109,460 | 6,914 | alexey-milovidov | @nikitamikhaylov You can merge after:
1. Get rid of try/catch.
2. Add an integration test. | 2019-09-13T05:59:10 | 2019-09-13T05:59:10 | {} | MEMBER |
ClickHouse | ClickHouse | 531,109,913 | 6,914 | maqroll | I'll do It.
@nikitamikhaylov is it ok for you? | 2019-09-13T06:01:30 | 2019-09-13T06:45:50 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,337,935 | 6,937 | amosbird | yeah, initializing it properly also fixes the bug. | 2019-09-13T18:11:30 | 2019-09-13T18:11:30 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,345,754 | 6,937 | amosbird | @akuzm you just caught me in the middle :) . That commit was supposed to run a CI test but I soon realize it's too slow so I did locally... | 2019-09-13T18:35:05 | 2019-09-13T18:35:05 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,541,399 | 6,880 | andyyzh | @yingfeng please add my Weixin: yangzhaohui168
Discuss a lot of this. | 2019-09-15T07:18:37 | 2019-09-15T07:18:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,727,024 | 6,938 | KochetovNicolai | I can't reproduce this because I don't know what is in `dist_cdp_cust_tag`. Can you create a test which is reproducible from scratch?
Also, can you please check if #6911 fixes your issue? | 2019-09-16T10:48:19 | 2019-09-16T10:48:19 | {} | MEMBER |
ClickHouse | ClickHouse | 530,711,178 | 6,898 | filimonov | 1) `<openSSL>` section in config.xml
2) how did you obtain certificates.
3) who connects to secure port? Only other clickhouse nodes?
4) does all clickhouse nodes have the same `<openSSL>` | 2019-09-12T07:55:37 | 2019-09-12T12:02:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,175,142 | 4,411 | filimonov | Let's continue tracking that issue in #1562 | 2019-09-13T09:49:12 | 2019-09-13T09:49:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,332,818 | 6,937 | amosbird | I guess it's because ZeroValueStorage is not initialized. https://github.com/yandex/ClickHouse/blob/master/dbms/src/Common/HashTable/HashTable.h#L223
. I'm not sure how memsan reasons about this failure, but I remember the ubsan error is due to the invalid pos = 0xffffffffffffffff and pos + field_size overflows. | 2019-09-13T17:55:43 | 2019-09-13T17:55:43 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,332,584 | 6,766 | dimarub2000 | [look here](https://github.com/yandex/ClickHouse/blob/e765733a26cfc4cecc13c981686560338256a6b1/dbms/src/Interpreters/convertFieldToType.cpp#L184) What could it mean?
I made a PR, will see how tests go. | 2019-09-13T17:54:59 | 2019-09-13T17:56:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,687,989 | 6,943 | filimonov | Execution time before exception:
```
18.14
1 rows in set. Elapsed: 0.716 sec.
19.11
1 rows in set. Elapsed: 19.155 sec.
``` | 2019-09-16T08:38:22 | 2019-09-16T08:38:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,702,719 | 6,917 | alexey-milovidov | We can unfold these macros at the table creation
but we should not unfold all other macros (shard, layer, etc.)
because there is typical scenario when table metadata (`.sql`) is copied between shards.
And the proposed logic of unfolding some macros and not unfolding other macros looks questionable. | 2019-09-16T09:26:54 | 2019-09-16T09:26:54 | {} | MEMBER |
ClickHouse | ClickHouse | 531,723,972 | 6,914 | maqroll | > @maqroll How things going? Do you need my help? Maybe it is better to reassign this task to me?
@nikitamikhaylov If it's ok I'll report the status at the end of the day (in Spain) and you'll decide. | 2019-09-16T10:38:12 | 2019-09-16T10:38:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,753,442 | 6,947 | akuzm | How much slower it is? I'd prefer to have some knob to turn it off.
Btw, tried setting up auto-building of indexes in gdb 8.3, and it didn't work for unknown reasons. | 2019-09-16T12:21:49 | 2019-09-16T12:21:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,838,406 | 6,922 | pachico | Interesting, it might be good to include it in the docs.
I'll try to find some time and PR it. | 2019-09-12T13:57:13 | 2019-09-12T13:57:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,867,634 | 5,417 | akuzm | > In master now? You mean it's introduced by the iterator inside out pr ?
I thought so, but after reverting this PR, it still reproduces. Will investigate further. | 2019-09-12T15:02:08 | 2019-09-12T15:02:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,960,526 | 5,417 | amosbird | > Interestingly, the test doesn't fail if I create the table and wait for the background merges to complete.
Intriguing. I'll investigate too. What's the cmake command line you use to generate memory sanitizer build? | 2019-09-12T18:59:11 | 2019-09-12T18:59:11 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,448,549 | 6,937 | amosbird | I cannot reproduce the thread san build failure locally using
```
docker run -it --rm --net=host -v /ci/tsandebs:/package_folder yandex/clickhouse-stateless-test
```
Is it some flaky test? | 2019-09-14T04:42:37 | 2019-09-14T04:42:37 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 530,818,890 | 6,922 | den-crane | duplicate https://github.com/yandex/ClickHouse/issues/1562 , https://github.com/yandex/ClickHouse/issues/4411 | 2019-09-12T13:12:09 | 2019-09-12T13:12:09 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,134,772 | 6,898 | ilyakarol | We've modified our SSL config a bit (increased number of connections, used more strong ciphers, etc.).
I am going to wait for a couple of days and try to catch this bug again.
Will post all info here. | 2019-09-13T07:40:52 | 2019-09-13T07:40:52 | {} | NONE |
ClickHouse | ClickHouse | 531,167,962 | 5,843 | filimonov | @nvartolomei not sure if there is a sense for making that constraint default.
Motivation:
1) "don't pay for what you don't use" principle
2) backward compatibility
What do you think? | 2019-09-13T09:26:05 | 2019-09-13T09:26:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,210,521 | 6,935 | filimonov | I've tried to disable `low_cardinality_allow_in_native_format`, even worse:
```
cat data_arrays.tsv | time -p clickhouse-client -q 'INSERT INTO test_arrays_lc FORMAT TSV' --log_queries=1 --query_id=insert_into_test_arrays_lc --max_insert_block_size=50000 --low_cardinality_allow_in_native_format=0
real 78.98
user... | 2019-09-13T12:03:52 | 2019-09-13T12:04:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,287,719 | 6,937 | amosbird | Actuall we can also make all StringRef zero value storages pointing to a global empty string. | 2019-09-13T15:40:44 | 2019-09-13T15:40:44 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,319,900 | 6,937 | akuzm | Great, thanks for investigating!
Your test case reproduces the error for me, we probably should add it to the tests. It has a shorter form using the new Values table function: `select groupUniqArray(v) from values('id int, v Array(int)', (1, [2]), (1, [])) group by id`.
So, is the core cause that we compare unini... | 2019-09-13T17:16:16 | 2019-09-13T17:22:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,710,362 | 6,903 | alexey-milovidov | Please try the new version and the setting and tell us the results. | 2019-09-16T09:50:37 | 2019-09-16T09:50:37 | {} | MEMBER |
ClickHouse | ClickHouse | 531,179,021 | 6,922 | filimonov | Closing that in advance to #1562. | 2019-09-13T10:02:15 | 2019-09-13T10:02:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,489,177 | 6,648 | alexey-milovidov | The first argument can be String, FixedString or Array. The second argument should be integer.
Both arguments can be constant / non-constant. | 2019-09-14T15:35:53 | 2019-09-14T15:35:53 | {} | MEMBER |
ClickHouse | ClickHouse | 531,559,541 | 6,938 | yuzhichang | Here's the bisect between v19.14.1.1104-testing(good) and v19.14.1.1112-testing(bad):
```
e986d6ba530e005cc3c2a6a5e84de1cdb9134c97 good
06bb0af3869b44ceff7ce4c69c6a23fa3bb4942f bad
```
I concluded that `LIVE VIEW` pr #6425 introduced this bug. | 2019-09-15T11:57:45 | 2019-09-15T11:57:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,795,654 | 6,920 | filimonov | You can't.
I.e. you will need to rewrite few files in the directory to make it possible.
You can look on that as a reference: https://github.com/yandex/ClickHouse/blob/master/utils/convert-month-partitioned-parts/main.cpp | 2019-09-12T12:08:44 | 2019-09-12T12:08:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,801,988 | 3,326 | filimonov | checked on 19.14 | 2019-09-12T12:28:21 | 2019-09-12T12:28:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,803,161 | 4,903 | filimonov | @kindred77 can you check 19.14? | 2019-09-12T12:32:10 | 2019-09-12T12:32:10 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,953,292 | 5,417 | akuzm | > > In master now? You mean it's introduced by the iterator inside out pr ?
>
> I thought so, but after reverting this PR, it still reproduces. Will investigate further.
It doesn't reproduce if I revert the key holder patch. The test that fails under MemorySanitizer is 00376_shard_group_uniq_array_of_int_array. I... | 2019-09-12T18:40:11 | 2019-09-12T18:40:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,174,774 | 2,496 | filimonov | Let's continue tracking that issue in #1562 | 2019-09-13T09:48:02 | 2019-09-13T09:48:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,333,256 | 6,937 | akuzm | Looking at emplaceIfZero, I don't see where the constructor is called for zero value Cell at all. Is this correct?
Before, there was an explicit initialization of zero StringRef in GroupUniqArray:
https://github.com/yandex/ClickHouse/pull/6729/files#diff-9281c6b3757a84abfb5643adf2bb48d2L251
but I removed it in t... | 2019-09-13T17:57:00 | 2019-09-13T17:57:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,478,679 | 6,648 | ucasfl | @alexey-milovidov | 2019-09-14T13:10:53 | 2019-09-14T13:10:53 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,709,841 | 6,914 | nikitamikhaylov | @maqroll How things going? Do you need my help? Maybe it is better to reassign this task to me? | 2019-09-16T09:48:59 | 2019-09-16T09:48:59 | {} | MEMBER |
ClickHouse | ClickHouse | 531,710,179 | 6,903 | alexey-milovidov | Performance of ORDER BY table's sorting key has been improved since version 19.14 if you enable the `optimize_read_in_order` setting. | 2019-09-16T09:50:04 | 2019-09-16T09:50:04 | {} | MEMBER |
ClickHouse | ClickHouse | 531,720,443 | 6,168 | nikitamikhaylov | @YiuRULE Tests are still red and I think the test *00975_move_partition_merge_tree* causes this.
To debug it locally (and do not fritter away your time on waiting tests in CI) you can run
```
./clickhouse-test 00975_move_partition_merge_tree
```
in *../ClickHouse/dbms/test/* directory. | 2019-09-16T10:25:23 | 2019-09-16T10:25:23 | {} | MEMBER |
ClickHouse | ClickHouse | 530,952,516 | 3,490 | den-crane | it could be anti-meltdown/spectre impact
https://make-linux-fast-again.com/
noibrs noibpb nopti nospectre_v2 nospectre_v1 l1tf=off nospec_store_bypass_disable no_stf_barrier mds=off mitigations=off | 2019-09-12T18:37:59 | 2019-09-12T18:37:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,007,292 | 6,931 | den-crane | There were an issue https://github.com/yandex/ClickHouse/issues/6308 19.11.4.24
release 19.11.7.40, 2019-08-14 : Fix infinite loop when reading Kafka messages. #6354
https://github.com/yandex/ClickHouse/blob/master/CHANGELOG.md | 2019-09-12T21:07:44 | 2019-09-12T21:07:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,282,086 | 5,417 | amosbird | @akuzm found the cause. It's actually a bug lurking before and gets unveiled by hashtable refactoring.
When getting data from `ColumnArray` https://github.com/yandex/ClickHouse/blob/master/dbms/src/Columns/ColumnArray.cpp#L144
we can get an invalid pointer when current array is empty. (due to left padding of podar... | 2019-09-13T15:25:18 | 2019-09-13T15:25:18 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,700,431 | 6,944 | alexey-milovidov | Ok.
PS. The test relies on assumption that tables are listed in alphabetical order. And this assumption is true (whenever it become false, the test will show it). | 2019-09-16T09:19:27 | 2019-09-16T09:19:27 | {} | MEMBER |
ClickHouse | ClickHouse | 531,712,987 | 6,871 | alexey-milovidov | For the reference:
`LRUCache became inconsistent` is most likely the result of memory corruption. Memory corruption may be caused by software or hardware fault. If software error is suspected, you should upgrade to the latest version. If hardware error is suspected, check the `dmesg`, `mcelog`, isolate the issue acr... | 2019-09-16T09:59:41 | 2019-09-16T09:59:41 | {} | MEMBER |
ClickHouse | ClickHouse | 530,982,405 | 5,417 | amosbird | Hmm, I cannot reproduce with memory sanitizer | 2019-09-12T19:59:50 | 2019-09-12T19:59:50 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,178,722 | 6,922 | filimonov | > BTW, `Ternary operator always calculate two branches` -- not a bug but a performance optimization.
But it still should be fixed someday - for example by passing stream with conditions checking mask to each branch.
| 2019-09-13T10:01:19 | 2019-09-13T10:01:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,218,625 | 6,924 | SaltTan | The reason seems to be: #5758 Extend KeyCondition interface so that it returns BoolMask. | 2019-09-13T12:31:05 | 2019-09-13T12:31:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,282,704 | 6,937 | amosbird | This bug can be reproduced using the following query
```
SELECT groupUniqArray(v)
FROM
(
SELECT
1 AS id,
[2] AS v
UNION ALL
SELECT
1 AS id,
[] AS v
)
GROUP BY id
```
clickhouse build command is
```
cmake -DSANITIZE=memory -DCMAKE_C_COMPILER="clang-8" -D... | 2019-09-13T15:26:58 | 2019-09-13T15:27:25 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 531,718,277 | 6,943 | nicelulu | > except that are in primary key
Maybe also need the columns used by all the ```Dependencies```. | 2019-09-16T10:17:53 | 2019-09-16T10:20:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,798,584 | 6,902 | abyss7 | Also will be backported to some stable versions. See future changelog for details. | 2019-09-12T12:17:21 | 2019-09-12T12:17:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,853,323 | 6,886 | nikitamikhaylov | Please add labels. | 2019-09-12T14:31:12 | 2019-09-12T14:31:12 | {} | MEMBER |
ClickHouse | ClickHouse | 530,857,729 | 5,417 | amosbird | In master now? You mean it's introduced by the iterator inside out pr ? | 2019-09-12T14:41:00 | 2019-09-12T14:41:00 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 530,888,927 | 6,886 | alexey-milovidov | @nikitamikhaylov Amos Bird cannot add labels by himself. | 2019-09-12T15:51:40 | 2019-09-12T15:51:40 | {} | MEMBER |
ClickHouse | ClickHouse | 530,933,356 | 6,168 | nikitamikhaylov | @YiuRULE Many tests were not successful because of heap-use-after-free error. You can click on *Details* button, scroll down and see *stderr.txt* file. Or I'll try to debug it ASAP. | 2019-09-12T17:48:42 | 2019-09-12T17:49:05 | {} | MEMBER |
ClickHouse | ClickHouse | 531,124,311 | 6,931 | jeffw-wherethebitsroam | @den-crane That sounds like the same bug, but I am running ClickHouse server version 19.13.1 revision 54425. | 2019-09-13T07:04:38 | 2019-09-13T07:04:38 | {} | NONE |
ClickHouse | ClickHouse | 531,173,691 | 6,914 | nikitamikhaylov | If you are interested in it, that's ok for me :) | 2019-09-13T09:44:20 | 2019-09-13T09:44:20 | {} | MEMBER |
ClickHouse | ClickHouse | 531,204,119 | 5,417 | akuzm | > > Interestingly, the test doesn't fail if I create the table and wait for the background merges to complete.
>
> Intriguing. I'll investigate too. What's the cmake command line you use to generate memory sanitizer build?
I'm using these flags:
```
cmake -DSANITIZE=memory -DCMAKE_C_COMPILER="clang-8" -DCMAKE_C... | 2019-09-13T11:38:24 | 2019-09-13T11:38:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,291,472 | 6,673 | proller | No.
```
$ ninja -j1 -v
[1/1761] : && /usr/bin/ccache /usr/bin/c++ -fdiagnostics-color=always -fsized-deallocation -pipe -msse4.1 -msse4.2 -mpopcnt -fno-omit-frame-pointer -Wall -Wnon-virtual-dtor -w -std=c++1z -O2 -g -DNDEBUG -O3 -fuse-ld=gold -rdynamic -Wl,--no-undefined -rdynamic contrib/protobuf/cmake/CM... | 2019-09-13T15:51:14 | 2019-09-13T15:51:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,489,366 | 6,907 | alexey-milovidov | @blinkov `documentation` tag prevents running code-related CI checks. | 2019-09-14T15:38:37 | 2019-09-14T15:38:37 | {} | MEMBER |
ClickHouse | ClickHouse | 531,696,381 | 6,943 | alexey-milovidov | It's related to compress buffer size and write buffer size. Both are 1 MB by default.
It looks like buffers for all columns are allocated at once, and 20 000 columns lead to about 40 GB of buffers. | 2019-09-16T09:06:05 | 2019-09-16T09:06:05 | {
"+1": 2
} | MEMBER |
ClickHouse | ClickHouse | 530,795,928 | 6,912 | ildus | @akuzm thank you, going to investigate the issue. Also could you tell why style checker doesn't like these two lines?
https://clickhouse-test-reports.s3.yandex.net/6912/20db1a75c6a241f6c55f56db9c337485022450fd/style_check.html | 2019-09-12T12:09:21 | 2019-09-12T12:09:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,344,069 | 6,941 | alexey-milovidov | Test needed. | 2019-09-13T18:29:48 | 2019-09-13T18:29:48 | {} | MEMBER |
ClickHouse | ClickHouse | 531,489,219 | 6,648 | alexey-milovidov | If you will have any difficulties, it's Ok to support only String data type for the first argument. | 2019-09-14T15:36:30 | 2019-09-14T15:36:30 | {} | MEMBER |
ClickHouse | ClickHouse | 531,708,057 | 6,904 | alexey-milovidov | It was fixed in recent versions.
The reason is that ALIAS columns were out of the common columns list. | 2019-09-16T09:43:16 | 2019-09-16T09:43:16 | {} | MEMBER |
ClickHouse | ClickHouse | 531,190,510 | 6,916 | abyss7 | It's an intended behavior due to the malformed message. You should recreate your Kafka table and add the setting `kafka_skip_broken_messages=N`. | 2019-09-13T10:45:17 | 2019-09-13T10:45:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,264,523 | 6,681 | vasyaabr | +1
Please fix it | 2019-09-13T14:42:24 | 2019-09-13T14:42:24 | {} | NONE |
ClickHouse | ClickHouse | 531,301,107 | 4,406 | renatga | I do migration from 19.9.2.4 to 19.13.3.26 Clickhouse version and continue see this error. Was it actually fixed? I was thinking to add the file "docker_related_config.xml" as you described upper, but as I see it is already there(the change is included to Dockerfile already). It is really annoing since those ports (812... | 2019-09-13T16:18:12 | 2019-09-13T16:56:25 | {} | NONE |
ClickHouse | ClickHouse | 531,710,744 | 6,943 | nicelulu | @alexey-milovidov Can be removed when inserting ```AddingDefaultBlockOutputStream```, because there may be 2000 columns, but only 100 column has a value. | 2019-09-16T09:52:05 | 2019-09-16T09:52:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,798,266 | 6,834 | abyss7 | Can you confirm that issue is fixed in 19.15 ? | 2019-09-12T12:16:14 | 2019-09-12T12:16:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,166,078 | 5,843 | filimonov | Since 19.14 and https://github.com/yandex/ClickHouse/pull/5273 :
```
CREATE TABLE test
(
`a` Int32,
`b` Enum8('b1' = 1, 'b2' = 2),
CONSTRAINT b_constraint CHECK b IN ('b1', 'b2')
)
ENGINE = MergeTree()
PARTITION BY tuple()
ORDER BY tuple()
Ok.
laptop5591.local :) INSERT INTO test (a,b) se... | 2019-09-13T09:20:19 | 2019-09-13T09:20:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,186,179 | 3,490 | lapkofear | We will update our production cluster next week and double check fix. | 2019-09-13T10:28:03 | 2019-09-13T10:28:03 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 531,715,613 | 6,943 | alexey-milovidov | @nicelulu That's great idea. We should not write unspecified columns at all (except that are in primary key). (They will be materialized in background merges, but if vertical merge algorithm is performed, it should not require too much memory.) | 2019-09-16T10:08:40 | 2019-09-16T10:08:40 | {} | MEMBER |
ClickHouse | ClickHouse | 531,744,964 | 6,937 | amosbird | Ok, how can I trigger a retest without force push new stuff? | 2019-09-16T11:52:51 | 2019-09-16T11:52:51 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 533,925,193 | 7,047 | alexey-milovidov | Looks trivial to fix. Just move `DISK_USAGE_COEFFICIENT_TO_SELECT` to `MergeTreeSettings`. | 2019-09-22T22:39:21 | 2019-09-22T22:39:21 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.