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 | 783,707,756 | 21,089 | cw9 | Ah I see, there is a new field `background_fetches_pool_size` that I didn't config, I had `background_schedule_pool_size = 16`.
Can someone share some insights on why the new config? | 2021-02-22T22:04:30 | 2021-02-22T22:04:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,713,608 | 21,089 | den-crane | It seems Yandex internally has a small number of Replicated tables and very fast and reliable network.
3 is enough for them and not for everybody else. | 2021-02-22T22:14:49 | 2021-02-22T22:14:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,806,817 | 19,873 | alanpaulkwan | Could we use DuckDB's Parquet reader (irrc is in turn from Impala which emphasizes Parquet)? If I recall correct from an email with Clickhouse-feedback circa 2019 you guys use parquetreader --> json but this has got to be inefficient?
The fact that the .json conversion is employed is probably why schema inference is... | 2021-02-23T01:51:36 | 2021-02-23T01:53:10 | {} | NONE |
ClickHouse | ClickHouse | 783,848,442 | 21,089 | cw9 | I did upgrade to 20.12.4 recently | 2021-02-23T03:20:44 | 2021-02-23T03:20:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,993,389 | 21,073 | kssenii | @Slach, it should not apply, because it is not initialised yet. Look https://clickhouse.tech/docs/en/sql-reference/statements/system/#query_language-system-reload-dictionaries | 2021-02-23T08:11:05 | 2021-02-23T08:11:05 | {} | MEMBER |
ClickHouse | ClickHouse | 784,419,691 | 21,112 | kssenii | @khlystov, how often these single row inserts are made? | 2021-02-23T18:34:21 | 2021-02-23T18:34:21 | {} | MEMBER |
ClickHouse | ClickHouse | 784,950,690 | 20,585 | alesapin | Just found that it's not a problem with StateMachine or LogStore. Very rarely for some reason commit inside nuraft requires several seconds:
```
2021.02.24 12:32:20.068311 [ 15934 ] {} <Debug> NuKeeperTCPHandler: Finished 785
2021.02.24 12:32:20.068445 [ 15991 ] {} <Debug> NuKeeperStateMachine: Processing request 78... | 2021-02-24T09:48:45 | 2021-02-24T09:59:43 | {} | MEMBER |
ClickHouse | ClickHouse | 784,962,605 | 20,585 | alesapin | Ahh 7 seconds is my operation timeout. | 2021-02-24T10:08:06 | 2021-02-24T10:08:06 | {} | MEMBER |
ClickHouse | ClickHouse | 783,623,068 | 20,946 | tavplubix | Stress test (memory) - `ALTER TABLE mt_00763_1 MODIFY column s String;` hung waiting for mutation | 2021-02-22T19:38:56 | 2021-02-22T19:38:56 | {} | MEMBER |
ClickHouse | ClickHouse | 783,720,917 | 21,089 | cw9 | so background_fetches_pool_size is related to number of tables? | 2021-02-22T22:27:08 | 2021-02-22T22:27:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,838,747 | 21,061 | liubo0127 | The MySQL example:
Here are two tables `one` and `two`, with the following requirements
```
DELETE
FROM one
LEFT JOIN two ON one.id=two.id
AND one.name=two.name
AND one.address=two.address
WHERE one.address IS NULL
```
How do I do it with ClickHouse | 2021-02-23T02:44:54 | 2021-02-23T02:44:54 | {} | NONE |
ClickHouse | ClickHouse | 783,856,895 | 21,091 | den-crane | duplicate https://github.com/ClickHouse/ClickHouse/issues/19857 | 2021-02-23T03:49:18 | 2021-02-23T03:49:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,859,604 | 17,089 | taiyang-li | @alesapin I am curious about the reproduce process of the bug you had fixed, can you provide it? | 2021-02-23T03:55:59 | 2021-02-23T03:55:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,353,147 | 3,413 | amirvaza | +1 | 2021-02-23T17:03:12 | 2021-02-23T17:03:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,721,915 | 21,115 | SE2AI | Thanks Denny! I check #8322 detailedly, it really helps a lot.
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?
If merge is resource expensive, what is `OPTIMIZE` designed for and what are the most appropr... | 2021-02-24T03:21:10 | 2021-02-24T03:32:59 | {} | NONE |
ClickHouse | ClickHouse | 783,421,481 | 21,041 | OmarBazaraa | > Need to check if ClickHouse at least tries to connect to this MongoDB instance.
>
> 1. Check the messages in logs. Also `SET send_logs_level = 'trace'` before executing the query in clickhouse-client.
> 2. Check tcpdump.
1. Now I start to get different error instead of timeout exception:
```sql
select * fr... | 2021-02-22T14:40:12 | 2021-02-22T14:40:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,574,376 | 14,893 | spearson23 | We also need this functionality. Is this being worked on? If not, I may be able to get someone to help. | 2021-02-22T18:21:55 | 2021-02-22T18:21:55 | {} | NONE |
ClickHouse | ClickHouse | 784,107,580 | 21,101 | filimonov | Duplicate of #13613, fixed in #19805.
Update to v20.12.6.29-stable or newer. | 2021-02-23T10:43:54 | 2021-02-23T10:43:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,116,660 | 20,946 | tavplubix | d5ea345aa9df9c77c95f9e5952251884407ae663:
Stress test (address) - `SYSTEM SYNC REPLICA` hung waiting on `StorageReplicatedMergeTree::waitForShrinkingQueueSize`
Stress test (thread) - `SYSTEM RELOAD DICTIONARIES` hung inside `LoadingDispatcher::loadImpl` | 2021-02-23T11:00:14 | 2021-02-23T11:00:14 | {} | MEMBER |
ClickHouse | ClickHouse | 784,420,372 | 21,112 | khlystov | @kssenii actually it happens exactly every 3 minutes
2021.02.23 18:31:29.387849 [ 21028 ] {} <Error> StorageRabbitMQ (test): Producer's channel 216 error: connection lost
2021.02.23 18:31:29.387996 [ 21028 ] {} <Error> StorageRabbitMQ (test): Library error report: connection lost
2021.02.23 18:34:30.193880 [ 210... | 2021-02-23T18:35:43 | 2021-02-23T18:35:43 | {} | NONE |
ClickHouse | ClickHouse | 784,511,428 | 21,106 | den-crane | Merges are eventual. They could NEVER happen.
Merge scheduler is not aware about GraphiteMergeTree reducing. It does NOT use rollup schema when planing the next merge.
Merge scheduler with 99% probability will not merge data older than M months.
> Isn't it should honour the conf from rollup.xml?
It's not imple... | 2021-02-23T21:04:04 | 2021-02-23T21:04:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,551,969 | 21,114 | dmitriy-myz | I suspect this https://github.com/ClickHouse/ClickHouse/pull/17397 | 2021-02-23T22:15:42 | 2021-02-23T22:15:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,706,877 | 21,096 | amosbird | > Automatic schema creation from clickhouse column description is another task.
Thanks for explaining, I understand now. The task is better defined in this way. | 2021-02-24T02:41:28 | 2021-02-24T02:41:28 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 784,824,729 | 21,119 | fastio | I'll add the test case soon. | 2021-02-24T06:28:52 | 2021-02-24T06:28:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,494,291 | 21,081 | den-crane | https://github.com/ClickHouse/ClickHouse/issues/21075 | 2021-02-22T16:22:32 | 2021-02-22T16:22:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,728,636 | 21,089 | den-crane | So you did not upgrade your CH ?
| 2021-02-22T22:41:23 | 2021-02-22T22:41:23 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,836,847 | 13,574 | mdianjun | @xjewer I have a question：
Theoretically, all parts of a partition will be merged into one, and maybe it will be very large. Then "MOVE PART" is not appropriate for that large part, since we might expect some of the data (not all data) in the partition to be moved to the new shard. | 2021-02-23T02:37:39 | 2021-02-23T02:37:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,865,016 | 21,073 | Slach | @kssenii yep, you are right,
LAYOUT(COMPLEX_KEY_HASHED())
works as expected in SQL
aslo
changed XML file, works fine only after apply ` SELECT dictGet('postgres_dict','service_type',tuple('mysql'));`
maybe it's expected behavior for lazy XML dictionaries loading
shall SYSTEM RELOAD DICTIONARIES apply for XM... | 2021-02-23T04:16:55 | 2021-02-23T04:17:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,363,871 | 19,291 | alexey-milovidov | Please check if this fixes #9900. | 2021-02-23T17:18:42 | 2021-02-23T17:18:42 | {} | MEMBER |
ClickHouse | ClickHouse | 784,562,252 | 21,103 | filimonov | It was added in 20.10.
See #14742. Check also: https://github.com/ClickHouse/ClickHouse/pull/18130
Please update to one of the latest versions | 2021-02-23T22:34:47 | 2021-02-23T22:34:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,566,875 | 21,061 | filimonov | Not sure if i understand it good. How `joinGet('join','id',id)` will help you?
Or you mean smth like `joinGet('join','id',(id,name,address))`?
The second definitely make sence. | 2021-02-23T22:44:28 | 2021-02-23T22:49:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,676,069 | 21,100 | zhangjmruc | Currently, the asof join allow to have one closest match condition, like>, >=, <, <=. The workaround for the condition t1.b != t2.b is to use UNION ALL for > and <.
select t1.a
,t1.b
,t2.c
from t1 **asof left join** t2
on t1.a = t2.a
and **t1.b < t2.b**
where t2.a = ''
**union all**
select t1.a
,t1.b
,t2.c... | 2021-02-24T01:57:18 | 2021-02-24T02:02:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,406,851 | 12,405 | alexey-milovidov | It also does not work when I try to convert types:
```
milovidov-desktop :) SELECT n, source FROM ( SELECT toDateTime64(number*1000, 3) AS n, 'original' AS source FROM numbers(10) WHERE number % 3 = 1 ) ORDER BY n WITH FILL STEP toDateTime64(1000, 3);
SELECT
n,
source
FROM
(
SELECT
... | 2021-02-22T14:17:47 | 2021-02-22T14:17:47 | {} | MEMBER |
ClickHouse | ClickHouse | 783,447,547 | 21,075 | den-crane | faster test:
```sql
CREATE TABLE e ( id String, time DateTime , m String )
ENGINE = MergeTree
PARTITION BY toYYYYMMDD(time)
ORDER BY (m, cityHash64(id))
SAMPLE BY cityHash64(id);
insert into e select toString(number),
toDateTime('2020-01-01 00:00:00') + number,
toString(number%3)
from number... | 2021-02-22T15:17:45 | 2021-02-22T15:17:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,496,753 | 21,080 | filimonov | You need to map arrow arrays to clickhouse arrays of corresponding subtypes, and create tests for that.
It should be smth like type detection / mapping + smth like 'fillColumnWithArrayData' template(?) function (maybe with some specialization will be needed for particular subtypes) + at least one test for every arra... | 2021-02-22T16:25:48 | 2021-02-22T16:25:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,501,230 | 12,571 | alexey-milovidov | Postponed until #21047. | 2021-02-22T16:32:20 | 2021-02-22T16:32:20 | {} | MEMBER |
ClickHouse | ClickHouse | 783,562,418 | 21,050 | azat | >@azat, test 01691_DateTime64_clamp fails in master
The behavior had been slightly changed in #21023 , and had been fixed already in 871f15343350fd1ba4e065c6f47672b3944b3319 | 2021-02-22T18:04:38 | 2021-02-22T18:04:38 | {} | MEMBER |
ClickHouse | ClickHouse | 783,586,007 | 21,020 | azat | >BTW query profiler can profile itself (for example Real time profiler can measure how frequently CPU time profiler was active). This is somewhat useful property.
I see, thanks. Okay, let's truncate the buffer size to 4K then. | 2021-02-22T18:39:56 | 2021-02-22T18:39:56 | {} | MEMBER |
ClickHouse | ClickHouse | 783,771,027 | 21,054 | cyberhuman | > 1. Why do you use Replicated Tables as temporary instead of MergeTree?
`ALTER TABLE table_source MOVE PARTITION partition_expr TO TABLE table_dest`
requires this:
> Both tables must be the same engine family (replicated or non-replicated).
https://clickhouse.tech/docs/en/sql-reference/statements/alter/par... | 2021-02-23T00:11:49 | 2021-02-23T00:11:49 | {} | NONE |
ClickHouse | ClickHouse | 783,846,546 | 20,219 | godliness | @yiguolei What if when we insert data into the vshard which is relocating to another node? | 2021-02-23T03:14:23 | 2021-02-23T03:17:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,838,990 | 21,078 | kssenii | BTW, `Docs check` should pass, but it doesn't | 2021-02-24T06:55:50 | 2021-02-24T06:55:50 | {} | MEMBER |
ClickHouse | ClickHouse | 784,904,491 | 21,119 | fastio | > You can pick tests from here if you want https://github.com/ClickHouse/ClickHouse/pull/20141/files
@nvartolomei Thanks. | 2021-02-24T08:34:26 | 2021-02-24T08:34:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,546,096 | 21,079 | azat | By the way the reason this wasn't captured by tests (but you can see that the test was executed too long, and [logs](https://clickhouse-test-reports.s3.yandex.net/16193/863c0992540c68b781b393a35d8c8f47dddbdd20/integration_tests_(thread)/integration_run_test_distributed_ddl_parallel_test_py_0.txt.out.log) contains excep... | 2021-02-22T17:38:39 | 2021-02-22T17:38:39 | {} | MEMBER |
ClickHouse | ClickHouse | 783,723,431 | 21,089 | den-crane | related to a number of simultaneous fetches.
if you make 1 insert per second into 1 table you need to fetch 1 part.
if you make 1 insert per second into 100 tables you need to fetch 100 parts. | 2021-02-22T22:31:34 | 2021-02-22T22:31:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,101,717 | 21,018 | filimonov | I've seen 2 similar reports already, which let me think that we store zookeerer ref somethere in some cases we use the obsolete one instead of the proper one from the singleton.
For example i had a report that
```
SELECT total_rows FROM system.tables WHERE table = 'Replicated table'
```
was returning "session... | 2021-02-23T10:33:44 | 2021-02-23T10:36:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,672,085 | 21,061 | liubo0127 | > Not sure if i understand it good. How `joinGet('join','id',id)` will help you?
> Or you mean smth like `joinGet('join','id',(id,name,address))`?
>
> The second definitely make sence.
Thanks, but how to create table with `joinGet('join','id',(id,name,address))` | 2021-02-24T01:46:35 | 2021-02-24T02:00:20 | {} | NONE |
ClickHouse | ClickHouse | 784,899,079 | 21,078 | Slach | @kssenii
how fix *.md files which generate links like following?
```
INFO:root:Tag <a href="#../../sql-reference/table-functions/jdbc/">JDBC</a>
INFO:root:Link to nowhere: #../../sql-reference/table-functions/jdbc/
INFO:root:Tag <a href="#../../../engines/table-engines/integrations/">Движки таблиц для интеграции<... | 2021-02-24T08:24:45 | 2021-02-24T08:24:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,607,656 | 14,893 | alexey-milovidov | @spearson23
As far as I know @RedClusive is also started to work on this task for his coursework.
But don't be afraid - this task has multiple loosely related parts and it can be tried by multiple developers concurrently. | 2021-02-22T19:14:23 | 2021-02-22T19:14:23 | {} | MEMBER |
ClickHouse | ClickHouse | 784,137,065 | 21,058 | windylcx | I follow the [docs] (https://clickhouse.tech/docs/en/development/build-osx/) | 2021-02-23T11:33:32 | 2021-02-23T11:34:00 | {} | NONE |
ClickHouse | ClickHouse | 784,363,451 | 17,868 | alexey-milovidov | No.
What will address #9900 is #19291. | 2021-02-23T17:18:03 | 2021-02-23T17:18:03 | {} | MEMBER |
ClickHouse | ClickHouse | 784,418,311 | 21,112 | khlystov | @kssenii from 1 to 3 minutes. just one row | 2021-02-23T18:32:10 | 2021-02-23T18:32:10 | {} | NONE |
ClickHouse | ClickHouse | 784,726,102 | 21,096 | BigRantLing | > So i think we still need possibility to put in the existing schema in textish format (which is more flexible, just to avoid neccecerity of direct filesystem access to clickhouse server filesystem for database users).
Thanks for explaining ! We use Clickhouse type to generate formate schema file to avoid type matc... | 2021-02-24T03:31:52 | 2021-02-24T03:31:52 | {} | NONE |
ClickHouse | ClickHouse | 784,807,522 | 16,669 | vasyaabr | Reproduced on ClickHouse server version 20.8.6 revision 54438, while making queries to database with MySQL engine | 2021-02-24T06:04:14 | 2021-02-24T06:04:14 | {} | NONE |
ClickHouse | ClickHouse | 783,412,791 | 21,067 | adrian17 | Reproducible. A tiny little bit better, but still orders of magnitude worse than in 21.1.x.
| WITHs in outer query | WITHs in inner query | time in 21.1.5.4 | time in 21.2.4.6 | time in 21.3.1.6083 |
| --- | --- | --- | --- | --- |
| 2 | 2 | 0.080s | 0.8s | 0.65s |
| 3 | 2 | 0.080s | 2.5s | 2.25s |
| 3 | 3 | 0.0... | 2021-02-22T14:26:45 | 2021-02-22T14:32:58 | {} | NONE |
ClickHouse | ClickHouse | 783,422,318 | 21,050 | tavplubix | @azat, test `01691_DateTime64_clamp` fails in master | 2021-02-22T14:41:25 | 2021-02-22T14:41:25 | {} | MEMBER |
ClickHouse | ClickHouse | 783,821,525 | 21,095 | zhangjmruc | For the config.xml with relative path in users_config, like <users_config>users.xml</users_config>:
Before fix:
-------------
~/app/clickhouse/clickhouse/bin$ clickhouse-local -c /home/frank/snowball-local-test/conf/config.xml -q "select * from table"
Processing configuration file '/home/frank/snowball-local-te... | 2021-02-23T02:20:29 | 2021-02-23T02:20:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,013,040 | 21,096 | BigRantLing | > > But why using ClickHouse type definition instead of passing concrete format in textish form?
>
> IMO it's intuitive to see the structured type of a format in order to use it for importing/exporting.
@amosbird @filimonov Thanks for suggesting. We don't use Protobuf syntax so that users could define the schema ... | 2021-02-23T08:51:23 | 2021-02-23T08:51:23 | {} | NONE |
ClickHouse | ClickHouse | 784,134,834 | 21,058 | windylcx | > Use command
> `ld -v`
> show the linker version.
@(#)PROGRAM:ld PROJECT:ld64-450.3
BUILD 18:16:53 Apr 5 2019
configured to support archs: armv6 armv7 armv7s arm64 arm64e arm64_32 i386 x86_64 x86_64h armv6m armv7k armv7m armv7em
LTO support using: LLVM version 10.0.1, (clang-1001.0.46.4) (static support for ... | 2021-02-23T11:29:39 | 2021-02-23T11:29:39 | {} | NONE |
ClickHouse | ClickHouse | 784,893,613 | 21,119 | nvartolomei | You can pick tests from here if you want https://github.com/ClickHouse/ClickHouse/pull/20141/files | 2021-02-24T08:15:12 | 2021-02-24T08:15:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,727,225 | 21,089 | cw9 | hmm, despite I have more tables recently than before, the inserts per second across all tables didn't change much, is there anything else that can impact data replication speed? Also is there anything else I should check other than system.replication_queue table? | 2021-02-22T22:38:53 | 2021-02-22T22:38:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,963,132 | 20,219 | yiguolei | > @yiguolei What if when we insert data into the vshard which is relocating to another node?
During relocating, one replica is added to the replica group. It is just like add a new replica in current **Shard**. And all data insert into the vshard will replicated to all replicas in the replica group. And the newly ad... | 2021-02-23T07:10:30 | 2021-02-23T07:12:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,001,486 | 20,159 | mzealey | I'm a bit curious as to why `optimize_read_in_order` doesnt fix this performance issue? | 2021-02-23T08:28:15 | 2021-02-23T08:28:15 | {} | NONE |
ClickHouse | ClickHouse | 784,709,022 | 21,054 | fastio | > I think it's because ClickHouse tries to drop metadata from the main ZK, not from auxiliary one:
> https://github.com/ClickHouse/ClickHouse/blob/2a36d6cb55af14b0dcf87c1b806afbf5c7dec8be/src/Storages/StorageReplicatedMergeTree.cpp#L753
It's a bug. | 2021-02-24T02:47:28 | 2021-02-24T02:47:28 | {
"+1": 3
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,421,548 | 20,087 | ervikrant06 | Thanks. It helps.
No merges I can see
~~~
time="2021-02-22 14:29:05 UTC" level=info msg="Merges will be applied: 0"
time="2021-02-22 14:29:05 UTC" level=info msg="Optimizations round is over, going to sleep for 1h0m0s"
~~~
Do you know which function is good for aggregation purpose? avg or any or something... | 2021-02-22T14:40:16 | 2021-02-22T14:40:16 | {} | NONE |
ClickHouse | ClickHouse | 783,576,073 | 21,033 | den-crane | WUT ?
>При работе со строками используйте `LowCardinality` вместо [Enum](enum.md).
почему "используйте"? Это не всегда и не всем подходит. | 2021-02-22T18:24:34 | 2021-02-22T18:24:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,793,632 | 21,054 | den-crane | > (state Committed). It is a bug.
ah, OK , it's a bug https://github.com/ClickHouse/ClickHouse/issues/14582 | 2021-02-23T01:10:53 | 2021-02-23T01:10:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,512,841 | 21,106 | den-crane | Basically GraphiteMergeTree is abandoned probably in favor of TTL group by. | 2021-02-23T21:06:35 | 2021-02-23T21:06:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,708,010 | 21,054 | fastio | > @fastio It's related to multiple ZooKeeper cluster, could you please take a look?
Hi @alexey-milovidov， I will take a look at this issue today. | 2021-02-24T02:44:24 | 2021-02-24T02:44:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 782,281,558 | 12,601 | azat | #16947 took too long in finding it path in upstream and now those patches does not work anymore, and not only #16947 does not work, but materialize trick does not helps either (for upstream):
```sql
select _shard_num, * from remote('127.{1,2}', system.one) limit 1 by dummy; -- fails
select materialize(_shard_num),... | 2021-02-19T19:09:03 | 2021-02-22T16:40:07 | {} | MEMBER |
ClickHouse | ClickHouse | 783,535,953 | 21,073 | kssenii | @Slach, see https://clickhouse.tech/docs/ru/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-structure/#ext_dict_structure-key.
A dictionary key can be either a UInt64 or a tuple.
You are using a string as a key and therefore there is an error of trying to convert a String to UInt64 and so a ... | 2021-02-22T17:23:11 | 2021-02-22T17:47:07 | {} | MEMBER |
ClickHouse | ClickHouse | 783,843,469 | 21,093 | bookeezhou | vim utils/check-mysql-binlog/main.cpp
add
DB::MySQLReplication::RowsEventHeader rows_header(header.type);
rows_header.parse(*event_payload);
two sentence in below：
case DB::MySQLReplication::WRITE_ROWS_EVENT_V1:
case DB::MySQLReplication::WRITE_ROWS_EVENT_V2:
{
... | 2021-02-23T03:03:15 | 2021-02-23T03:03:15 | {} | NONE |
ClickHouse | ClickHouse | 783,987,991 | 21,084 | filimonov | Loosely related to https://github.com/ClickHouse/ClickHouse/issues/19228 | 2021-02-23T08:00:08 | 2021-02-23T08:00:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,270,410 | 17,868 | filimonov | Will it close https://github.com/ClickHouse/ClickHouse/pull/9900 ? | 2021-02-23T15:10:36 | 2021-02-23T15:10:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,416,713 | 21,112 | kssenii | @khlystov, what is the average time period between two `connection lost` logs? And how much data was approximately inserted between them? | 2021-02-23T18:29:37 | 2021-02-23T18:35:05 | {} | MEMBER |
ClickHouse | ClickHouse | 784,421,116 | 21,112 | khlystov | @kssenii just once. I am prototyping right now and do such single row inserts for tests | 2021-02-23T18:37:03 | 2021-02-23T18:37:03 | {} | NONE |
ClickHouse | ClickHouse | 783,493,616 | 21,080 | filimonov | Great news! Nobody is currently working on the task so it's yours.
See also: https://github.com/ClickHouse/ClickHouse/issues/7704
Please let us know if you will need any help / support with that, and keep us up to date. | 2021-02-22T16:21:39 | 2021-02-22T16:28:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,829,824 | 14,893 | pingyu | I'm still working on this task, and I'm sorry for the late...
I will speed up. | 2021-02-23T02:24:31 | 2021-02-23T02:24:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,843,606 | 21,093 | bookeezhou | this issue can close | 2021-02-23T03:03:42 | 2021-02-23T03:03:42 | {} | NONE |
ClickHouse | ClickHouse | 784,020,374 | 21,096 | BigRantLing | > 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
Thank you for reply ！
In the first version of this issue , we do act... | 2021-02-23T09:05:28 | 2021-02-23T09:05:51 | {} | NONE |
ClickHouse | ClickHouse | 784,702,251 | 21,098 | amosbird | Tests failures are unrelated. | 2021-02-24T02:28:53 | 2021-02-24T02:28:53 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 784,843,945 | 20,585 | alesapin | Tests Ok, but I don't understand why we have extremely rare `Connection loss`. Will try to investigate. | 2021-02-24T07:04:34 | 2021-02-24T07:04:34 | {} | MEMBER |
ClickHouse | ClickHouse | 783,496,197 | 21,075 | den-crane | it was not ideal in 20.3
```sql
SELECT count() FROM e SAMPLE 2 / 100 WHERE m = '1'
1 rows in set. Elapsed: 0.020 sec. Processed 98.30 thousand rows
SELECT count() FROM e SAMPLE 2 / 100 WHERE m IN ('1');
Elapsed: 0.017 sec. Processed 196.61 thousand rows
SELECT count() FROM e SAMPLE 2 / 100 WHERE m IN (SELE... | 2021-02-22T16:25:07 | 2021-02-22T16:25:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,500,764 | 21,054 | alexey-milovidov | @fastio It's related to multiple ZooKeeper cluster, could you please take a look? | 2021-02-22T16:31:37 | 2021-02-22T16:31:37 | {} | MEMBER |
ClickHouse | ClickHouse | 783,566,103 | 21,015 | azat | >The question is - how to improve our CI to detect these kind of issues automatically.
Indeed, by some reason ASan does not detect this leak, it may be related to google/sanitizers#937
I've found one case when gcc-10 ASan does find leak, while clang-11/12 doesn't:
```c
#include <cstdlib>
int main(int argc,... | 2021-02-22T18:10:24 | 2021-02-22T18:10:47 | {} | MEMBER |
ClickHouse | ClickHouse | 784,175,864 | 20,585 | alesapin | Functional stateless tests (release, DatabaseReplicated)
`Cannot allocate block number in ZooKeeper: Coordination::Exception: Connection loss` :(
```
2021.02.23 13:04:56.479313 [ 158 ] {} <Trace> NuKeeperTCPHandler: Received heartbeat for session #0
2021.02.23 13:05:09.248701 [ 158 ] {} <Information> NuKeeperTCPH... | 2021-02-23T12:45:12 | 2021-02-23T12:45:58 | {} | MEMBER |
ClickHouse | ClickHouse | 784,358,373 | 16,481 | amosbird | Test failure looks unrelated https://clickhouse-test-reports.s3.yandex.net/16481/605dd9a971d72c91a721148e0c489c08dd4f86c7/integration_tests_(thread)/integration_run_test_storage_kafka_test_py_1.txt.out.log | 2021-02-23T17:10:37 | 2021-02-23T17:10:37 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 783,501,651 | 21,061 | filimonov | There was a PR for iptrie dicts recently. https://github.com/ClickHouse/ClickHouse/issues/18241
BTW: why do you need it? Normally you pass the `id` to `joinGet` call. | 2021-02-22T16:33:01 | 2021-02-22T16:36:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,969,874 | 21,096 | amosbird | 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. | 2021-02-23T07:23:25 | 2021-02-23T07:24:42 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 783,502,646 | 21,075 | artbeglaryan | When you put some non existing values in `IN`, reads less again.
```
SELECT count()
FROM e
SAMPLE 2 / 100
WHERE m IN ('1', '10000')
┌─count()─┐
│ 19529 │
└─────────┘
1 rows in set. Elapsed: 0.010 sec. Processed 1.13 million rows, 31.35 MB (108.94 million rows/s., 3.02 GB/s.)
SELECT count()
FROM e
S... | 2021-02-22T16:34:28 | 2021-02-22T16:34:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,555,637 | 21,080 | den-crane | 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:arra... | 2021-02-22T17:53:56 | 2021-02-22T18:06:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,583,393 | 21,033 | kitaisreal | @den-crane https://github.com/ClickHouse/ClickHouse/pull/21086. | 2021-02-22T18:35:56 | 2021-02-22T18:35:56 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 783,675,481 | 21,060 | tavplubix | `Fast test — fail: 1` - it was broken in master, let's restart tests | 2021-02-22T21:08:10 | 2021-02-22T21:08:10 | {} | MEMBER |
ClickHouse | ClickHouse | 783,985,469 | 21,096 | filimonov | Sounds interesting. But why using ClickHouse type definition instead of passing concrete format in textish form?
Why it may be simpler / safer:
1) sometimes the ClickHouse types can be ambiguous in the terms of concrete format. Typical example: Boolean vs UInt8. Mapping of Nested may also be very confusing.
2) we do... | 2021-02-23T07:56:41 | 2021-02-23T07:58:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,000,156 | 21,096 | amosbird | > But why using ClickHouse type definition instead of passing concrete format in textish form?
IMO it's intuitive to see the structured type of a format in order to use it for importing/exporting. | 2021-02-23T08:25:39 | 2021-02-23T08:25:39 | {
"+1": 2,
"confused": 1
} | COLLABORATOR |
ClickHouse | ClickHouse | 784,205,315 | 21,087 | vzakaznikov | This PR is ready for review and merge. | 2021-02-23T13:32:31 | 2021-02-23T13:32:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 784,554,729 | 21,096 | filimonov | >
>
> > But why using ClickHouse type definition instead of passing concrete format in textish form?
>
> IMO it's intuitive to see the structured type of a format in order to use it for importing/exporting.
Intuitive - yes. But also limiting (in protobuf there can be sub messages, used one inside another, com... | 2021-02-23T22:20:39 | 2021-02-23T22:38:10 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 774,959,910 | 20,179 | alex-zaitsev | @excitoon will look into that | 2021-02-08T08:16:04 | 2021-02-08T08:16:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 774,960,778 | 10,043 | filimonov | See similar in PMM:
https://www.percona.com/blog/2020/06/23/extending-percona-monitoring-and-management-for-mysql-innodb-cluster-with-custom-queries/
https://github.com/percona/mysqld_exporter/blob/master/collector/custom_query.go
There are several (currently 3) folders for resolutions (high-resolution, med, lo... | 2021-02-08T08:17:41 | 2021-02-08T08:17:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 774,480,593 | 19,961 | Avogar | >It would be better to do it via virtual column instead of a setting. Currently, _shard_num virtual column is returned in SELECTs from distributed tables, so the natural extension is to use it in INSERTs in order to write to a specific shard.
@alexey-milovidov What do you think? Current implementation LGTM. | 2021-02-06T13:50:54 | 2021-02-08T11:27:10 | {} | MEMBER |
ClickHouse | ClickHouse | 775,134,997 | 16,424 | alexey-milovidov | #20192 | 2021-02-08T13:02:36 | 2021-02-08T13:02:36 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.