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 | 526,107,262 | 6,720 | qietingfengling | PARTITION BY date ORDER BY (timestamp-timestamp%3600000, service) TTL date + INTERVAL 1 DAY;
this is my schema. My date value is "2019-12-12", timestamp value is 1567070700000,My table segment type is date、String、float64 | 2019-08-29T09:35:01 | 2019-08-29T09:35:01 | {} | NONE |
ClickHouse | ClickHouse | 526,138,805 | 6,721 | helloboy1234 | OK，I'll try GCC-9. Thanks!
------------------ 原始邮件 ------------------
发件人: "alexey-milovidov"<notifications@github.com>;
发送时间: 2019年8月29日(星期四) 晚上7:05
收件人: "yandex/ClickHouse"<ClickHouse@noreply.github.com>;
抄送: "晴天"<408996957@qq.com>;"Mention"<mention@noreply.github.com>;
主题: Re: [yandex/ClickHouse] Ubant... | 2019-08-29T11:09:05 | 2019-08-29T11:09:05 | {} | NONE |
ClickHouse | ClickHouse | 526,142,418 | 3,470 | demenkov | > Now it returns `2106-02-07 06:28:16`. Even better.
I think it is not better, because when use in interval it will produce unexpected results.
`SELECT
close_time,
count(*)
FROM mt4_trade
WHERE close_time <= '1970-01-01 00:00:00'
GROUP BY close_time
ORDER BY close_time ASC
LIMIT 10
┌─────────... | 2019-08-29T11:21:43 | 2019-08-29T11:21:43 | {} | NONE |
ClickHouse | ClickHouse | 526,351,870 | 6,707 | jcdang | You're right... I wrote it in python this time.
https://gist.github.com/jcdang/b9a42a4976c389b2259202f8af264095
➜ perf time python3 keepalive.py
python3 keepalive.py 0.99s user 0.07s system 6% cpu 16.180 total
➜ perf time ./nokeepalive.sh
./nokeepalive.sh 1.93s user 1.79s system 18% cpu 20.595 total
➜ perf ... | 2019-08-29T20:38:29 | 2019-08-29T20:38:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,353,576 | 4,355 | nvartolomei | This seems to be fixed at least in:
```
SELECT *
FROM system.build_options
WHERE name = 'VERSION_DESCRIBE'
┌─name─────────────┬─value──────────────┐
│ VERSION_DESCRIBE │ v19.13.2.19-stable │
└──────────────────┴────────────────────┘
``` | 2019-08-29T20:43:39 | 2019-08-29T20:43:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,367,698 | 6,707 | filimonov | OK. Now let's get back to the question: do we really need such a feature?
I understand that it's quite handy, and actually i like the idea (for example it's cool for tests), and for me the concept of "garbage in / garbage out" is clean. But I almost sure that if it will be available the users will misuse that, and t... | 2019-08-29T21:22:26 | 2019-08-29T21:22:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,385,451 | 6,707 | jcdang | We have someone in our company that absolutely loves ClickHouse performance and uses clickhouse-client extensively. He has written planning code in SQL that operates sequentially on the data. It's come up several times -- he asks why is our application not as fast as when he runs it with clickhouse-client locally. W... | 2019-08-29T22:27:28 | 2019-08-29T22:27:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,241,905 | 6,414 | abyss7 | Should be fixed in 19.14+ | 2019-08-29T15:36:41 | 2019-08-29T15:36:41 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,242,366 | 6,549 | alexey-milovidov | Does the issue reproduce if you add
`<min_merge_bytes_to_use_direct_io>0</min_merge_bytes_to_use_direct_io>`
into `<merge_tree>` section in config.xml:
```
<merge_tree>
<min_merge_bytes_to_use_direct_io>0</min_merge_bytes_to_use_direct_io>
</merge_tree>
```
Don't forget to uncomment `<merge_tree>` sec... | 2019-08-29T15:37:43 | 2019-08-29T15:37:43 | {} | MEMBER |
ClickHouse | ClickHouse | 526,315,277 | 6,594 | vmarkovtsev | @alesapin
```sql
ATTACH TABLE uasts
(
`id` Int32,
`repo` String,
`lang` String,
`file` String,
`line` Int32,
`parents` Array(Int32),
`pkey` String,
`roles` Array(Int16),
`type` String,
`uptypes` Array(String),
`value` String
)
ENGINE = MergeTree()
... | 2019-08-29T18:52:05 | 2019-08-29T18:52:05 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 526,375,663 | 6,714 | filimonov | > After the change it worked fine, I am not sure this could damage the 'player' table on the cluster. By the way server-2 is the replica with the issues)
In certain circumstances it definitely can make a mess in zookeeper. I'm not 100% sure but i would say with some 'extra luck' it may be a source of the problem you... | 2019-08-29T21:49:28 | 2019-08-29T21:50:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,395,737 | 6,735 | volfco |
I've also run more complex queries and the processed row numbers also fluctuate like so:
```
cluster :) select toStartOfHour(Timestamp) as H, count(*) from metrics.histograms_dist group by H order by H;
SELECT
toStartOfHour(Timestamp) AS H,
count(*)
FROM metrics.histograms_dist
GROUP BY H
ORDE... | 2019-08-29T23:16:04 | 2019-08-29T23:16:04 | {} | NONE |
ClickHouse | ClickHouse | 526,473,193 | 6,735 | filimonov | Can you show your <remote_servers> configuration? | 2019-08-30T06:17:37 | 2019-08-30T06:17:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 525,958,156 | 6,716 | den-crane | BTW,
>PARTITION BY substring(symbol,1,1)
>ORDER BY (symbol,date_time)
I hope your select looks like select .... from ... where symbol =
you don't need substring(symbol,1,1) in where. It will reduce performance. | 2019-08-28T23:16:31 | 2019-08-28T23:16:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,192,782 | 6,723 | alesapin | Failed tests are known and fail in master. | 2019-08-29T13:47:03 | 2019-08-29T13:47:03 | {} | MEMBER |
ClickHouse | ClickHouse | 526,352,954 | 6,707 | filimonov | > You're right... I wrote it in python this time.
Still not fair. Python has completely another overhead. You're should rewrite all 3 to python in that case.
```
>time python -c "print 'Hello, world'"
Hello, world
real 0m0,014s
user 0m0,014s
sys 0m0,000s
>time echo 'Hello, world'
Hello, world
real 0m... | 2019-08-29T20:41:48 | 2019-08-29T20:54:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,384,176 | 6,734 | SaltTan | Yes, this one works, thank you
```
SELECT alias AS name
FROM
(
SELECT name AS alias
FROM system.settings
) a
ANY INNER JOIN
(
SELECT name
FROM system.settings
) b USING (name)
WHERE name = 'enable_optimize_predicate_expression'
SETTINGS enable_optimize_predicate_expression = 1
``` | 2019-08-29T22:22:18 | 2019-08-29T22:22:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,585,196 | 6,549 | alexey-milovidov | > -r-xr-xr-x 1 clickhouse clickhouse 48 Aug 29 09:58 /data/clickhouse/data/actions/actions/20190708_786_786_0_263702/quality.mrk
The issue might be somewhat related (but should not) to FREEZE partitions. | 2019-08-30T12:37:08 | 2019-08-30T12:37:08 | {} | MEMBER |
ClickHouse | ClickHouse | 526,042,011 | 6,710 | filimonov | It's a bit contoversional as give no feedback per query, and behaviour of http status and headers related to progress / summary is not clear.
I don't think it will have any performance advantages comparing to multiple queries in single connection in one http keepalive session.
So i afraid that users will tend to ... | 2019-08-29T06:20:53 | 2019-08-29T06:21:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,436,118 | 6,721 | helloboy1234 | I still failed to compile with gcc-9,the error is the same as above.
------------------ 原始邮件 ------------------
发件人: "alexey-milovidov"<notifications@github.com>;
发送时间: 2019年8月29日(星期四) 晚上7:05
收件人: "yandex/ClickHouse"<ClickHouse@noreply.github.com>;
抄送: "晴天"<408996957@qq.com>;"Mention"<mention@noreply.githu... | 2019-08-30T02:55:23 | 2019-08-30T02:55:23 | {} | NONE |
ClickHouse | ClickHouse | 526,095,557 | 6,714 | filimonov | It's nice to get a feedback from such a chaos testing session. :)
General answer is: to replace one replica you don't need to recreate tables or do any destructive operations. But to make it work you need to have same server configuration and healthy zookeeper state. If you have zookeeper data damaged or server conf... | 2019-08-29T09:02:31 | 2019-08-29T09:02:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,143,100 | 6,594 | alesapin | https://github.com/yandex/ClickHouse/pull/6723 fix other issue. | 2019-08-29T11:24:07 | 2019-08-29T11:24:21 | {} | MEMBER |
ClickHouse | ClickHouse | 526,171,265 | 6,723 | alesapin | bug fix for 19.14 + | 2019-08-29T12:53:35 | 2019-08-29T12:53:35 | {} | MEMBER |
ClickHouse | ClickHouse | 526,242,892 | 4,736 | abyss7 | Closing for now, since there were no other reports. | 2019-08-29T15:38:54 | 2019-08-29T15:38:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,313,778 | 6,733 | 4ertus2 | #4974 | 2019-08-29T18:48:02 | 2019-08-29T18:48:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,324,659 | 6,707 | filimonov | That test is wrong. You don't use keealive, as curl is called for each query separately. | 2019-08-29T19:19:20 | 2019-08-29T19:19:20 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,372,985 | 6,728 | filimonov | On 19.11.8 - proper exception:
```
Code: 44, e.displayText() = DB::Exception: You can not modify columns used in index. Index name: 'idx' bad column: 'i64' (version 19.11.8.46 (official build))
``` | 2019-08-29T21:40:13 | 2019-08-29T21:40:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,477,474 | 6,549 | thyn | file access writes before mutation
-r-xr-xr-x 1 clickhouse clickhouse 48 Aug 29 09:58 /data/clickhouse/data/actions/actions/20190708_786_786_0_263702/quality.mrk
| 2019-08-30T06:34:27 | 2019-08-30T06:34:27 | {} | NONE |
ClickHouse | ClickHouse | 526,379,971 | 6,710 | filimonov | Please check the discussion in the #6707 issue before merging.
Maybe some less attractive name for the flag / feature should be chosen? | 2019-08-29T22:04:39 | 2019-08-30T08:10:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,539,623 | 6,736 | alesapin | Already fixed here https://github.com/yandex/ClickHouse/pull/6723 | 2019-08-30T09:49:35 | 2019-08-30T09:49:35 | {} | MEMBER |
ClickHouse | ClickHouse | 526,550,992 | 6,721 | alexey-milovidov | You still can use clang (versions 9, 8, 7 will work). | 2019-08-30T10:27:33 | 2019-08-30T10:27:33 | {} | MEMBER |
ClickHouse | ClickHouse | 526,064,627 | 6,720 | filimonov | @BayoNet can we add that to FAQ section of documentation? | 2019-08-29T07:35:06 | 2019-08-29T07:35:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,104,421 | 6,707 | jcdang | We are already using Keep-Alive.
@filimonov I hope you change your mind :)
I for one will find it dramatically beneficial. | 2019-08-29T09:27:03 | 2019-08-29T09:27:03 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,324,012 | 6,734 | 4ertus2 | It works if you add aliases for subqueries. We are going to force use of aliases for the subqueries with multiple tables in FROM section. Related issue and setting #4974 #6733 | 2019-08-29T19:17:22 | 2019-08-29T19:17:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,447,573 | 6,698 | inv2004 | ```
$ getent ahosts localhost
::1 STREAM localhost
::1 DGRAM
::1 RAW
127.0.0.1 STREAM
127.0.0.1 DGRAM
127.0.0.1 RAW
$ getent hosts localhost
::1 localhost localhost.localdomain localhost6 localhost6.localdomain6
```
```
clickhouse-client //... | 2019-08-30T04:05:32 | 2019-08-30T04:05:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,474,009 | 6,549 | thyn | > Does the issue reproduce if you add
>
> `<min_merge_bytes_to_use_direct_io>0</min_merge_bytes_to_use_direct_io>`
> into `<merge_tree>` section in config.xml:
>
> ```
> <merge_tree>
> <min_merge_bytes_to_use_direct_io>0</min_merge_bytes_to_use_direct_io>
> </merge_tree>
> ```
>
> Don't forget to unco... | 2019-08-30T06:20:58 | 2019-08-30T06:20:58 | {} | NONE |
ClickHouse | ClickHouse | 526,481,381 | 6,539 | George3d6 | @4ertus2 That's the approach I ended up using since it was fastest (SUM over a sub select with group by), but I feel like implementing sumDistinct, even with the same performance, might feel more coherent. Then again, maybe it's harder than I imagine to do something like this.
If I ever have the time I'll try lookin... | 2019-08-30T06:49:44 | 2019-08-30T06:49:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,151,194 | 6,720 | filimonov | How many insert per sec do you do? Do you use bulk inserts?
What can you see in `system.merges` table? | 2019-08-29T11:52:13 | 2019-08-29T11:52:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,234,317 | 6,587 | thyn | also
select top 100 createdon from tables where createdon>'20190724'
results contains 2019-07-15 00:00:00
| 2019-08-29T15:20:07 | 2019-08-29T15:20:07 | {} | NONE |
ClickHouse | ClickHouse | 526,236,803 | 6,717 | alexey-milovidov | `FATAL: ThreadSanitizer CHECK failed: /build/llvm-toolchain-8-8.0.1~svn363027/projects/compiler-rt/lib/sanitizer_common/sanitizer_deadlock_detector.h:69 "((n_all_locks_)) < (((sizeof(all_locks_with_contexts_)/sizeof((all_locks_with_contexts_)[0]))))" (0x40, 0x40)`
- known issue: https://github.com/google/sanitizers/is... | 2019-08-29T15:25:31 | 2019-08-29T15:25:31 | {} | MEMBER |
ClickHouse | ClickHouse | 526,254,247 | 4,736 | borismol | I also noticed that kafka engine lost data approximately 1%. If create another kafka engine and read data from beginning offsets again all data is written to table untli lag is not disappeared. Once lag is disapeared kafka engine lost data again.
Error log contains many
2019.08.29 13:20:27.778190 [ 33 ] {} <Trace> S... | 2019-08-29T16:05:52 | 2019-08-29T16:05:52 | {} | NONE |
ClickHouse | ClickHouse | 526,378,073 | 6,719 | filimonov | Please check `windowFunnel` and `sequenceMatch` functions. They are quite straightforward.
Also there is bit outdated but still nice video https://www.youtube.com/watch?v=YpurT78U2qA
(Outdated because most of it is about workaround of limitation of `sequenceMatch` which is not exist anymore since beginning of 201... | 2019-08-29T21:57:55 | 2019-08-29T21:57:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,556,108 | 6,741 | filimonov | /cc @hczhcz | 2019-08-30T10:47:07 | 2019-08-30T10:47:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,557,992 | 6,594 | alesapin | Reproduced on my machine. | 2019-08-30T10:54:25 | 2019-08-30T10:54:25 | {
"hooray": 2
} | MEMBER |
ClickHouse | ClickHouse | 526,044,481 | 6,710 | filimonov | Also lot of clients like jdbc / odbc get an base url as a connection conguration. If that base url will contain `multiquery` it will lead to unpredicable results. | 2019-08-29T06:30:18 | 2019-08-29T06:30:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,233,931 | 6,587 | thyn | I've updated YYYYMMDD is ok, but "YYYY-MM-DD" not. It's strange
select top 100 createdon from table where createdon>'2019-07-24'
Code: 41, e.displayText() = DB::Exception: Cannot parse datetime 2019-07-24 (version 19.13.2.19)
| 2019-08-29T15:19:15 | 2019-08-29T15:19:15 | {} | NONE |
ClickHouse | ClickHouse | 526,472,132 | 6,549 | thyn | 2. ls: cannot access /data/clickhouse/data/actions/actions/tmp_mut_20190708_786_786_0_263703/quality.mrk: No such file or directory
3. not yet
4. original file in 20190708_786_786_0_263703/quality.mrk is 48 byte size , quality.bin 14kb.
5. MergeTree
with some obfuscation. Clickhouse is standalone
` sql
CREA... | 2019-08-30T06:13:21 | 2019-08-30T06:29:08 | {} | NONE |
ClickHouse | ClickHouse | 526,483,977 | 6,734 | 4ertus2 | Issue is not solved yet. It’s just a workaround. Reopen. | 2019-08-30T06:59:35 | 2019-08-30T07:00:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,567,265 | 6,742 | alexey-milovidov | BTW, we cannot support any delimiter that may happen to be parsed as part of value.
For example `.` is not supported as a delimiter, because we cannot determine, what `1.2` is.
But this is another issue: we support padding by spaces and skip spaces around values unconditionally* (before parsing delimiter). Better t... | 2019-08-30T11:31:12 | 2019-08-30T11:32:18 | {} | MEMBER |
ClickHouse | ClickHouse | 526,037,476 | 6,667 | amosbird | Not sure how the POPULATE flag works. But via memory page touching, pre fault can be parallelized | 2019-08-29T06:02:31 | 2019-08-29T06:02:31 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 526,129,713 | 6,713 | alesapin | Completely incorrect. | 2019-08-29T10:39:37 | 2019-08-29T10:39:37 | {} | MEMBER |
ClickHouse | ClickHouse | 526,139,226 | 6,721 | helloboy1234 | My compiler version is low?
------------------ 原始邮件 ------------------
发件人: "akuzm"<notifications@github.com>;
发送时间: 2019年8月29日(星期四) 晚上6:39
收件人: "yandex/ClickHouse"<ClickHouse@noreply.github.com>;
抄送: "晴天"<408996957@qq.com>;"Author"<author@noreply.github.com>;
主题: Re: [yandex/ClickHouse] Ubantu build clic... | 2019-08-29T11:10:29 | 2019-08-29T11:10:29 | {} | NONE |
ClickHouse | ClickHouse | 526,152,629 | 6,722 | filimonov | Duplicate of #6480
Let's continue discussion there. | 2019-08-29T11:57:01 | 2019-08-29T11:57:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,187,920 | 6,726 | den-crane | ```
create table XX (A Int64, B String, Status UInt8) Engine = Memory;
insert into XX values(1, 'Status=1', 1), (2, 'Status=0', 0), (2, 'Status=2', 2), (3, 'Status=2', 2);
select * from XX
┌─A─┬─B────────┬─Status─┐
│ 1 │ Status=1 │ 1 │
│ 2 │ Status=0 │ 0 │
│ 2 │ Status=2 │ 2 │
│ 3 │ Status=2 │ ... | 2019-08-29T13:35:35 | 2019-08-29T13:35:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,190,216 | 6,707 | filimonov | Also (since you don't care about the result and want to decrease the roundtrip time) - please use `format Null` for your queries to make comparison more reliable. | 2019-08-29T13:40:52 | 2019-08-29T13:41:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,353,443 | 6,686 | vitlibar | Fixed in master. | 2019-08-29T20:43:14 | 2019-08-29T20:43:14 | {} | MEMBER |
ClickHouse | ClickHouse | 526,371,119 | 6,731 | filimonov | So you put a messages in Native format in the Kafka? How many rows in single message?
Can you check if names & order of columns you put in the Kafka Native stream matches exactly the create statement of Kafka table? | 2019-08-29T21:33:48 | 2019-08-29T21:34:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,039,535 | 6,710 | alex-krash | This functionality is inconsistent, cause it allows to send queries with different output format, and with different type (select/insert/alter). This will lead to trash output. | 2019-08-29T06:11:02 | 2019-08-29T06:11:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,238,302 | 6,549 | alexey-milovidov | @golubev-ml Could you please provide the stack trace of this exception? It is logged below. | 2019-08-29T15:28:49 | 2019-08-29T15:29:00 | {} | MEMBER |
ClickHouse | ClickHouse | 526,265,576 | 6,336 | igor-sh8 | We created our own service that reads data from Kafka and inserts it into the Clickhouse table with Null Engine. And in such schema Clickhouse works more efficiently (it can process much more data than table with Kafka Engine).
So maybe like a feature it is possible to add for example setting to Kafka Engine table wh... | 2019-08-29T16:36:30 | 2019-08-29T16:36:30 | {} | NONE |
ClickHouse | ClickHouse | 526,352,169 | 6,707 | filimonov | I've fixed your test to use keepelive (in bash / curl it's a bit tricky, but doable).
https://gist.github.com/filimonov/9b7a9d3d39e0446c9b4ad8db8dab6d29
I was checking localhost, and get the following results
```
> time bash nokeepalive.sh
real 0m2.482s
user 0m0.693s
sys 0m0.679s
>time bash keepalive.sh
r... | 2019-08-29T20:39:24 | 2019-08-29T20:39:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,357,400 | 6,707 | jcdang | That's just my shell setup showing the folder name. Sorry for the confusion.
Here is the results with your updates running across data centers with ClickHouse on a bigger box:
```
➜ perfmod time bash keepalive.sh
bash keepalive.sh 0.03s user 0.03s system 1% cpu 4.278 total
➜ perfmod time bash nokeepalive.sh
... | 2019-08-29T20:54:41 | 2019-08-29T20:54:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,042,293 | 3,490 | lapkofear | Unfortunately I can't currently check this fix due to hardware lack | 2019-08-29T06:22:02 | 2019-08-29T06:22:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,042,464 | 6,710 | filimonov | @jcdang did you try http keepalive as an alternative? | 2019-08-29T06:22:45 | 2019-08-29T06:22:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,129,535 | 6,721 | akuzm | 'internal compiler error' means that it is a compiler bug. It should be reported to gcc according to their guidelines: https://www.gnu.org/software/gcc/bugs/
The only issue with the same error message I could find should be already fixed in gcc 8.3: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84853 | 2019-08-29T10:39:04 | 2019-08-29T10:39:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,234,539 | 6,710 | kalexmills | Seems like a beneficial improvement to the code base for handling batch requests more efficiently. What would maintainers need to see in order to accept this PR? | 2019-08-29T15:20:35 | 2019-08-29T15:20:35 | {} | NONE |
ClickHouse | ClickHouse | 526,308,545 | 6,732 | alexey-milovidov | 1. `cat /proc/sys/vm/overcommit_memory`
2. Do you use Docker, cgroups or specifically modified ulimits? | 2019-08-29T18:33:44 | 2019-08-29T18:33:44 | {} | MEMBER |
ClickHouse | ClickHouse | 526,324,322 | 6,711 | dimarub2000 | Still fails. ( | 2019-08-29T19:18:18 | 2019-08-29T19:18:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,560,546 | 6,741 | hczhcz | ```
select arrayReduce('min', arrayPopBack([1])) = 0
select arrayReduce('avg', arrayPopBack([1])) = nan
```
Currently, `min()` is implemented like this. | 2019-08-30T11:04:48 | 2019-08-30T11:04:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,593,246 | 6,740 | alexey-milovidov | Just in case: deadlock cannot happen, because POSIX rwlock put all read locks in front of write lock:
http://man7.org/linux/man-pages/man3/pthread_rwlockattr_setkind_np.3.html | 2019-08-30T13:02:15 | 2019-08-30T13:02:15 | {} | MEMBER |
ClickHouse | ClickHouse | 526,073,073 | 6,710 | jcdang | @alex-krash Agreed. Garbage in, garbage out.
@filimonov The use case that I'm trying to handle is where there are ~350 of processing queries that need to be ran sequentially and I don't care about the output until the end. At the end I would ideally make a separate request to get the output. A keepalive session w... | 2019-08-29T08:00:31 | 2019-08-29T08:00:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,138,625 | 6,710 | jcdang | @filimonov Does this alleviate your connection configuration concern? | 2019-08-29T11:08:29 | 2019-08-29T11:08:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,227,636 | 6,707 | jcdang | This is without the format null
https://gist.github.com/jcdang/8a9242c6f4980d1e048f046651bb7146
➜ perf time ./keepalive.sh
./keepalive.sh 0.64s user 0.58s system 15% cpu 7.938 total
➜ perf time ./nokeepalive.sh
./nokeepalive.sh 0.64s user 0.58s system 14% cpu 8.202 total
➜ perf time ./multiquery.sh
./mult... | 2019-08-29T15:04:58 | 2019-08-29T15:04:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,356,635 | 4,355 | nvartolomei | Was fixed in https://github.com/yandex/ClickHouse/pull/4310/ | 2019-08-29T20:52:36 | 2019-08-29T20:52:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,387,695 | 6,714 | MeteHanC | > >
> I think that restart itself without all that manipulation would lead to the same result. The thing that helped migration to be finished - is disappearing of that malfunctioning node which didn't send any acknowledgements about the mutation state.
Hmm, I see. You are probably right. I deleted all the disks an... | 2019-08-29T22:37:29 | 2019-08-29T22:37:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 525,970,301 | 6,694 | heyciao | I have just tested everything with version 19.13.1.11-r3 (Gentoo Linux) and with this version everything looks good :) => closing this issue. | 2019-08-29T00:14:32 | 2019-08-29T00:14:32 | {} | NONE |
ClickHouse | ClickHouse | 526,057,615 | 208 | filimonov | It seems like https://github.com/yandex/ClickHouse/pull/5116 released in 19.8.3.8 should fix original issue by introducing `X-ClickHouse-Summary` header and improving `X-ClickHouse-Progress` behaviour.
Response buffering should allow to solve the problem of sending headers after data https://clickhouse.yandex/do... | 2019-08-29T07:14:02 | 2019-08-29T07:14:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,147,474 | 5,029 | letterkey | > seems there is no such way parse something like this
> `[{"a": 1, "b": 2}, {"a":1, "b":3}]`
>
> @alexey-milovidov shoule we have some function like this
>
> ```
> select JSONExtractArrayRaw('[{"a": 1, "b": 2}, {"a":1, "b":3}]')
> =>
> Array of Raw String
> ['{"a": 1, "b": 2}', '{"a":1, "b":3}']
> ```
t... | 2019-08-29T11:39:22 | 2019-08-29T11:39:22 | {} | NONE |
ClickHouse | ClickHouse | 526,146,410 | 6,549 | filimonov | Several questions:
1) Can you run `strace -f -e trace=file` and reproduce the problem.
2) Can you show the state of filesystem after that error? i.e. `ls -la /var/lib/clickhouse/path/to/problematic/part`
3) do you use docker?
4) can you check the size of problematic part? Is it bigger than 10Gb?
5) What table e... | 2019-08-29T11:35:54 | 2019-08-29T11:50:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,151,409 | 6,594 | WestDragon | Need to recreate the index? | 2019-08-29T11:52:56 | 2019-08-29T11:52:56 | {} | NONE |
ClickHouse | ClickHouse | 526,546,984 | 6,738 | akuzm | Use parentheses: `Array(String)`.
If you have further questions about arrays, it may be helpful to refer to the documentation: https://clickhouse.yandex/docs/en/data_types/array/ | 2019-08-30T10:13:24 | 2019-08-30T10:13:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,578,953 | 6,549 | filimonov | > But I see that I have more then 20 processes of clickhouse server.
Please recheck. It should be one process (and lot of threads).
| 2019-08-30T12:16:13 | 2019-08-30T12:16:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 525,959,877 | 6,716 | ghost | @den-crane yes the query is of the form:
```
select * from time_table where symbol='BLAH'
``` | 2019-08-28T23:24:13 | 2019-08-28T23:24:13 | {} | NONE |
ClickHouse | ClickHouse | 526,121,605 | 6,200 | CurtizJ | Implemented in #6274 | 2019-08-29T10:15:16 | 2019-08-29T10:15:16 | {
"rocket": 2
} | MEMBER |
ClickHouse | ClickHouse | 526,169,623 | 6,707 | filimonov | Since you already have the code - can you do benchmark to illustrate the difference and Performance improvement?
Let's say 100 queries sequence 1) w/o keepalive, 2) with keepalive, and 3) in batch/multiquery mode?
| 2019-08-29T12:49:16 | 2019-08-29T12:49:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,397,246 | 6,735 | volfco | Upgrading to `19.13.3.26` does not seem to resolve this issue. | 2019-08-29T23:23:58 | 2019-08-29T23:23:58 | {} | NONE |
ClickHouse | ClickHouse | 526,489,639 | 6,549 | thyn | > 1. Can you run `strace -f -e trace=file` and reproduce the problem.
Could you please advice how to correctly trace it. I can attach to pid. But I see that I have more then 20 processes of clickhouse server. Should I run clickhouse-server using PROG subcommand? What command line args should I pass to start clickhou... | 2019-08-30T07:20:29 | 2019-08-30T07:20:29 | {} | NONE |
ClickHouse | ClickHouse | 526,095,717 | 1,684 | MeteHanC | So what is the best way to create a cold replica? I just want to create a replica to keep the data but I don't want CH to query that replica? For a workaround I can write a script to make daily insertions manually to the cold replica or enable & disable clickhouse(which is Configured as a replica as well) daily to ge... | 2019-08-29T09:02:57 | 2019-08-29T09:02:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,173,399 | 6,594 | alesapin | @vmarkovtsev Can you provide table definition: `metadata/dbname/uasts.sql`? | 2019-08-29T12:58:55 | 2019-08-29T12:58:55 | {} | MEMBER |
ClickHouse | ClickHouse | 526,213,082 | 6,728 | alesapin | Without `idx` everything is ok. | 2019-08-29T14:31:51 | 2019-08-29T14:31:51 | {} | MEMBER |
ClickHouse | ClickHouse | 526,500,610 | 6,731 | igor-sh8 | Yes, from our custom service we insert data in Native format to the Kafka. In single message we have 3 rows (our tests).
For generating insert order we get metadata from the end table `db.testnativetable` in which order is the same as order in table with Kafka Engine. We get order of columns from such query (Insert... | 2019-08-30T07:56:46 | 2019-08-30T07:56:46 | {} | NONE |
ClickHouse | ClickHouse | 526,587,721 | 5,376 | akuzm | I will close this for now, feel free to reopen if there are any news. | 2019-08-30T12:45:28 | 2019-08-30T12:45:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,131,629 | 6,714 | MeteHanC | > It's nice to get a feedback from such a chaos testing session. :)
>
> From your story the most unclear moments are:
>
> 1. what do you mean by 're-initializing zookeeper instances'
> 2. what do you mean by 'It was looking good at that point' (at least you need to check system.replicas table after each step)
>... | 2019-08-29T10:45:47 | 2019-08-29T10:45:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,137,462 | 6,721 | alexey-milovidov | @helloboy1234 You can use gcc-9. | 2019-08-29T11:04:49 | 2019-08-29T11:04:49 | {} | MEMBER |
ClickHouse | ClickHouse | 526,221,132 | 6,729 | akuzm | @amosbird @alexey-milovidov @KochetovNicolai I'll be glad to hear your thoughts on this. | 2019-08-29T14:49:45 | 2019-08-29T14:49:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,263,516 | 6,729 | akuzm | > Will the compiler be smart enough to get rid of all the potential copies when constructing KeyHolders?
Hope so, but didn't check yet. If we see extra copying, we can change the interface slightly: pass keys as usual, and only pass wrappers when we have temporary StringRef keys. Then, make persistKey()/discardKey()... | 2019-08-29T16:30:30 | 2019-08-29T16:30:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,426,545 | 6,720 | qietingfengling | I have 50 nodes, each node inserts 2 or less batches per second, and one batch has 20W data. Some nodes have several data in the system.merges table, some have no data. | 2019-08-30T02:05:14 | 2019-08-30T02:07:48 | {} | NONE |
ClickHouse | ClickHouse | 526,550,738 | 6,698 | alexey-milovidov | Confirmed. This is known issue: only single DNS record is used. The issue exists from the beginning. | 2019-08-30T10:26:40 | 2019-08-30T10:26:40 | {} | MEMBER |
ClickHouse | ClickHouse | 526,565,477 | 5,417 | akuzm | Some benchmark results.
I took this PR, and also my refactored version from #6243. Rebased them both to `v19.13.1.1060-testing`. Then ran string_hash_map test, term48, 50M elements. The results are as follows:
|Branch |Hash map type |Emplaced keys/second|
|---|---|--- ... | 2019-08-30T11:24:37 | 2019-08-30T11:24:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 533,107,359 | 6,981 | abyss7 | It's a normal situation, that you can't build code with CMake that you don't have on a filesystem - run `git submodule update --init` before running CMake. | 2019-09-19T12:29:24 | 2019-09-19T12:29:24 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.