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 | 806,205,106 | 22,068 | filimonov | > against the Distributed Table node,
what is the type of query? Isn't it a `select count()` query? (may relate to #16108 )
> I am using ClickHouse 20.7.3.7
Can you check one of the latest releases? | 2021-03-24T21:43:07 | 2021-03-24T21:50:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,385,300 | 21,834 | alexey-milovidov | CC @qoega | 2021-03-24T00:46:52 | 2021-03-24T00:46:52 | {} | MEMBER |
ClickHouse | ClickHouse | 805,789,632 | 22,080 | Aureliano-Berlindia | Just found out that this problem is caused by keepalived
We try to use keepalived to achieve some kind of "HA",
and there are three clickhouse-servers:
```
192.168.1.1
192.168.1.2
192.168.1.3
```
by using keepalived , we set a vip 192.168.1.200, every request to 192.168.1.200 will redirect to the machine that... | 2021-03-24T12:43:56 | 2021-03-24T12:43:56 | {} | NONE |
ClickHouse | ClickHouse | 806,107,887 | 22,062 | alexey-milovidov | #22006 | 2021-03-24T19:36:14 | 2021-03-24T19:36:14 | {} | MEMBER |
ClickHouse | ClickHouse | 806,199,150 | 22,101 | filimonov | That is wrong. Quotes are needed.
```
ERROR_LOG_PATH=
ERROR_LOG_DIR="$(dirname $ERROR_LOG_PATH || true)"
dirname: missing operand
Try 'dirname --help' for more information.
```
With
```
set -eo pipefail
```
added in #18954 that error (missing operand) leads to exiting of the entrypoint.
Check altern... | 2021-03-24T21:30:41 | 2021-03-24T21:30:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,587,772 | 22,022 | magvan | > @magvan
> Получается что null надо приводить к 1900-01-01 и делать это через вьюху в PG. Какой то костыль получается.
С вьюхой тоже не получилось. Создал вьюху, теперь она отдает такие данные:
Попробов... | 2021-03-24T07:57:25 | 2021-03-24T08:01:32 | {} | NONE |
ClickHouse | ClickHouse | 806,203,937 | 22,067 | filimonov | similar to https://github.com/ClickHouse/ClickHouse/issues/22046 | 2021-03-24T21:40:41 | 2021-03-24T21:40:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,712,391 | 22,022 | filimonov | > Code: 36, e.displayText() = DB::Exception: Code: 36, e.displayText() = DB::Exception: Dictionary structure type of 'range_min' and 'range_max' should be an integer, Date, DateTime, or Enum. Actual 'range_min' and 'range_max' type is Nullable(Date) (version 21.3.3.14 (official build)) (version 21.3.3.14 (official buil... | 2021-03-24T10:46:30 | 2021-03-24T10:52:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,905,128 | 21,827 | krebrikov | Thanks to Denny Crane, who provided an insight to the root cause of problem.
There is a DateTime field **wdt** in ReplicatedMergeTree table of my scheme that was subjected to "Code 40: Checksums..." error
`Code: 40, ... Checksums of parts don't match: uncompressed hash of compressed files doesn't match (version 20.... | 2021-03-24T15:10:58 | 2021-03-24T15:10:58 | {} | NONE |
ClickHouse | ClickHouse | 806,039,293 | 22,068 | jlazdw | We have select_sequential_consistency = 0 and insert_quorum =2. Before I shut down the ZK cluster for the experiment, I already shutdown the data injection path, so insert_quorum should not matter.
I used clickhouse-benchmark against the Distributed Table node, on a small cluster (8 shards). For query latency < ... | 2021-03-24T17:58:52 | 2021-03-24T17:58:52 | {} | NONE |
ClickHouse | ClickHouse | 805,351,659 | 21,907 | alexey-milovidov | @abyss7 Better not to limit at all or limit with some large value (will be implicitly limited by max_memory_usage).
BTW, why do we need
```
/// Chunk is not completely in buffer, copy it to scratch space.
memory.resize(chunk_size);
```
Maybe it's possible to read data by parts that will fit in buffer and coun... | 2021-03-23T23:38:47 | 2021-03-23T23:38:47 | {} | MEMBER |
ClickHouse | ClickHouse | 805,546,576 | 21,837 | Jokser | @alexey-milovidov Your comments are addressed. Could you please review the change again? | 2021-03-24T06:45:18 | 2021-03-24T06:45:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,698,180 | 22,022 | filimonov | > Получается что null надо приводить к 1900-01-01 и делать это через вьюху в PG. Какой то костыль получается.
Не надо. 1900-01-01 это ошибка в документации, такая дата сейчкас для КХ невозможна. Там семерка потерялась. Должно быть как раз 1970-01-01
> Я ожидал что вернет 72.
Да, если вы всё верно сделали то ... | 2021-03-24T10:33:56 | 2021-03-24T10:33:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,741,007 | 22,068 | filimonov | Do you use any non-standard configuration for the table? select_sequential_consistency? insert_quorum?
How the queries you benchmark look like? How do you benchmark? (do you use clickhouse-benchmark?)
Does it reproduces always (so when you start zookeeper back the problem disappear?) | 2021-03-24T11:20:16 | 2021-03-24T11:20:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,817,040 | 22,022 | den-crane | https://github.com/ClickHouse/ClickHouse/issues/21647 | 2021-03-24T13:24:35 | 2021-03-24T13:24:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,927,840 | 21,948 | kitaisreal | @azat I thought about moving it under parameter initially, but this code breaks abstraction IDictionary IDictionarySource abstractions. I plan to put each source and dictionary in separate .cpp file and organize work between then only through interfaces.
Also there can be cases where this preallocation actually dec... | 2021-03-24T15:35:05 | 2021-03-24T15:35:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,195,759 | 22,100 | naylinn69 | Attempted fix here #22101 | 2021-03-24T21:24:19 | 2021-03-24T21:24:19 | {} | NONE |
ClickHouse | ClickHouse | 806,200,555 | 22,101 | naylinn69 | Hmm I managed to start the server with this change since the dirname cmd is ORed with true. Either way your fix looks better. Thanks for the quick fix! | 2021-03-24T21:33:35 | 2021-03-24T21:33:35 | {} | NONE |
ClickHouse | ClickHouse | 805,347,355 | 22,048 | alexey-milovidov | @Akazz Is it possibly be related to the recent changes in MySQL connection pools? | 2021-03-23T23:27:27 | 2021-03-23T23:27:27 | {} | MEMBER |
ClickHouse | ClickHouse | 805,367,820 | 22,017 | den-crane | incorrect indeed
```
19.13.7.57
SELECT
(NULL, NULL) = (8, 0),
(NULL, NULL) = (3, 2),
(NULL, NULL) = (0, 0),
(NULL, NULL) = (3, 1)
┌─equals(tuple(NULL, NULL), tuple(8, 0))─┬─equals(tuple(NULL, NULL), tuple(3, 2))─┬─equals(tuple(NULL, NULL), tuple(0, 0))─┬─equals(tuple(NULL, NULL), tuple(3, 1)... | 2021-03-24T00:17:05 | 2021-03-24T00:17:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,385,237 | 21,834 | alexey-milovidov | It makes sense.
Maybe we can define the boundaries with CONSTRAINTs in a table.
If all the constraints are in form of comparison, e.g. `column >= const`, we can analyze them for the boundaries of data to generate.
| 2021-03-24T00:46:41 | 2021-03-24T00:46:41 | {} | MEMBER |
ClickHouse | ClickHouse | 805,598,022 | 12,701 | yishanchuan | Hi, guys. I have a similar exception, but I use MergeTree! Following is the detail stack:
```
ru.yandex.clickhouse.except.ClickHouseException: ClickHouse exception, code: 33, host: 10.58.92.49, port: 8123; Code: 33, e.displayText() = DB::Exception: Cannot read all data. Bytes read: 1. Bytes expected: 5. (version 20... | 2021-03-24T08:15:13 | 2021-03-24T08:15:52 | {} | NONE |
ClickHouse | ClickHouse | 805,603,059 | 22,021 | magvan | > @den-crane
> ээээ, ну так и почему вы ожидаете что это должно работать?
> каким образом timestamp в PG сконвертируется в числовой тип в КХ?
Сделал получение получение словаря через вьюху в PG, что бы integer возвращался.
```
CREATE VIEW pay.v_rate AS
select
id,
currencies,
rate,
extract(e... | 2021-03-24T08:24:02 | 2021-03-24T08:26:12 | {} | NONE |
ClickHouse | ClickHouse | 805,740,003 | 22,022 | magvan | @filimonov
еще раз все проверил и запустил, заработало. Результат 72. Спасибо большое! | 2021-03-24T11:18:56 | 2021-03-24T11:18:56 | {} | NONE |
ClickHouse | ClickHouse | 805,933,491 | 21,948 | azat | >Also there can be cases where this preallocation actually decrease performance
Can you show me an example?
> or does not affect it at all for non sparse hash table
Agree, AFAIR initially it was done only for sparse hash tables, but later, during discussion, it was made more generic.
> If you have cases whe... | 2021-03-24T15:41:50 | 2021-03-24T15:42:10 | {} | MEMBER |
ClickHouse | ClickHouse | 806,152,070 | 22,094 | alexey-milovidov | @qoega Similar but different. I'll take this. | 2021-03-24T20:04:54 | 2021-03-24T20:04:54 | {} | MEMBER |
ClickHouse | ClickHouse | 806,158,337 | 20,213 | MaxJPowers | @nikitamikhaylov Any updates?
| 2021-03-24T20:16:27 | 2021-03-24T20:16:36 | {} | NONE |
ClickHouse | ClickHouse | 806,162,946 | 18,742 | alexey-milovidov | The task appeared to be too hard to implement correctly for condvars with timedwait functions expecting absolute time. | 2021-03-24T20:24:13 | 2021-03-24T20:24:13 | {} | MEMBER |
ClickHouse | ClickHouse | 805,354,577 | 21,902 | alexey-milovidov | PS. Version 20.10 is no longer supported and you should upgrade to 21.3. | 2021-03-23T23:46:38 | 2021-03-23T23:46:38 | {} | MEMBER |
ClickHouse | ClickHouse | 805,491,717 | 12,483 | zeekling | Thanks . | 2021-03-24T04:47:46 | 2021-03-24T04:47:46 | {} | NONE |
ClickHouse | ClickHouse | 805,617,241 | 21,774 | yangyongyongyong | > Do you copy & paste your long query in clickhouse-client interactive session? It may be for example some issue with tabs.
>
> Workarounds:
>
> * you can pipe in the file
> * use queries-file command line argument.
> * remove newlines and tabs from the query before pasting it into interactive session
error | 2021-03-24T08:48:52 | 2021-03-24T08:48:52 | {} | NONE |
ClickHouse | ClickHouse | 806,155,016 | 22,000 | alexey-milovidov | #22094. | 2021-03-24T20:10:34 | 2021-03-24T20:10:34 | {} | MEMBER |
ClickHouse | ClickHouse | 806,352,945 | 21,946 | azat | @akuzm can you take a look? (since you was merging original PR) | 2021-03-25T04:23:30 | 2021-03-25T04:23:30 | {} | MEMBER |
ClickHouse | ClickHouse | 805,533,444 | 22,021 | magvan | > ээээ, ну так и почему вы ожидаете что это должно работать?
> каким образом timestamp в PG сконвертируется в числовой тип в КХ?
судя по словарю в clickhouse он правильно сконвертировал
Дата из PG соо... | 2021-03-24T06:16:52 | 2021-03-24T06:16:52 | {} | NONE |
ClickHouse | ClickHouse | 805,641,280 | 21,639 | yanhuqing666 | Hi, I found a mistake in the Known issues:
```
contrib/jemalloc/include/jemalloc/internal/tsd_generic.h:145
assert(!malloc_slow && tsd_tcache_enabled_get(tsd) &&
tsd_reentrancy_level_get(tsd) == 0);
```
It should be
```
contrib/jemalloc/include/jemalloc/internal/tsd_generic.h:145
... | 2021-03-24T09:26:56 | 2021-03-24T09:26:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,899,684 | 22,086 | akuzm | I am somewhat confused about which operations we support. The basic block is a data part, that is immutable. The table consists of a set of data parts. There are many tables. The following operations on the table are possible:
1. Data part is added.
2. Several data parts are merged into one: source parts are dele... | 2021-03-24T15:06:17 | 2021-03-24T15:06:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,138,227 | 19,129 | alexey-milovidov | No longer relevant. Fixed. | 2021-03-24T19:54:25 | 2021-03-24T19:54:25 | {} | MEMBER |
ClickHouse | ClickHouse | 806,161,350 | 19,473 | alexey-milovidov | The bug was fixed in another PR. | 2021-03-24T20:21:30 | 2021-03-24T20:21:30 | {} | MEMBER |
ClickHouse | ClickHouse | 806,367,743 | 22,079 | sundy-li | continue in #21943 cc @azat | 2021-03-25T05:09:13 | 2021-03-25T05:09:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,409,147 | 21,770 | filimonov | @taylor12805 do you need help? | 2021-03-25T06:52:32 | 2021-03-25T06:52:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,342,584 | 22,022 | alexey-milovidov | @kitaisreal Наверное Nullable уже работает корректно для `range_hashed` и задачу можно закрывать? | 2021-03-23T23:15:47 | 2021-03-23T23:15:47 | {} | MEMBER |
ClickHouse | ClickHouse | 805,351,937 | 21,907 | alexey-milovidov | Let's also make the error message more descriptive. It should say about the chunk of HTTP Transfer-Encoding: chunked. | 2021-03-23T23:39:37 | 2021-03-23T23:39:37 | {} | MEMBER |
ClickHouse | ClickHouse | 805,358,604 | 22,017 | alexey-milovidov | According to the example of @den-crane, the behaviour of old versions is definitely incorrect.
```
SELECT ((NULL, NULL) = (8, 0)) OR ((NULL, NULL) = (3, 2)) OR ((NULL, NULL) = (0, 0)) OR ((NULL, NULL) = (3, 1))
```
How it can return 1 if all conditions are false?
```
SELECT
(NULL, NULL) = (8, 0),
(N... | 2021-03-23T23:56:44 | 2021-03-23T23:57:25 | {} | MEMBER |
ClickHouse | ClickHouse | 805,373,591 | 21,897 | alexey-milovidov | This is an obvious bug, #11844 | 2021-03-24T00:26:36 | 2021-03-24T00:27:29 | {} | MEMBER |
ClickHouse | ClickHouse | 805,568,296 | 21,574 | Sallery-X | > hi @filimonov,
>
> I implemented a node management feature. The node is stateful. If one replica node fails, the node can be paused by admin, and the user DDL operation can skip the paused node. When the node is restored, the node can be restarted by admin, and the DDL can continue to execute.
>
> I don’t know ... | 2021-03-24T07:24:14 | 2021-03-24T07:24:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,644,635 | 21,677 | alesapin | > Wow, something new
>
> ```
> INFO [2021-03-23 17:36:27,930] jepsen test runner - jepsen.core {:perf {:latency-graph {:valid? true},
> :rate-graph {:valid? true},
> :valid? true},
> :workload {:ok-count 156,
> :valid? false,
> :lost-count 3,
> :lost "#{231... | 2021-03-24T09:30:57 | 2021-03-24T09:30:57 | {} | MEMBER |
ClickHouse | ClickHouse | 805,937,140 | 21,948 | kitaisreal | Example benchmark with our HashTable
```
#include <iostream>
#include <iomanip>
#include <city.h>
#include <Core/Types.h>
#include <Common/Stopwatch.h>
#include <Common/HashTable/HashMap.h>
#include "config_core.h"
namespace DB
{
NO_INLINE void variant1(size_t max_size)
{
std::cerr << "HashesTest r... | 2021-03-24T15:46:01 | 2021-03-24T15:46:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,328,941 | 22,106 | wangzhen11aaa | May be LEFT ANY JOIN is the right solution. | 2021-03-25T03:04:39 | 2021-03-25T03:04:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,438,243 | 22,111 | filimonov | I expect it to work. Did you try that? | 2021-03-25T07:44:42 | 2021-03-25T07:44:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,416,429 | 21,574 | marising | hi @filimonov,
I implemented a node management feature. The node is stateful. If one replica node fails, the node can be paused by admin, and the user DDL operation can skip the paused node. When the node is restored, the node can be restarted by admin, and the DDL can continue to execute.
I don’t know if this... | 2021-03-24T02:02:35 | 2021-03-24T02:02:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,597,012 | 21,677 | alesapin | > ```
> 2021.03.23 18:35:24.119499 [ 30503 ] {} <Fatal> Application: Child process was terminated by signal 11.
> 2021.03.23 18:35:32.056719 [ 30743 ] {} <Fatal> BaseDaemon: ########################################
> 2021.03.23 18:35:32.056885 [ 30743 ] {} <Fatal> BaseDaemon: (version 21.4.1.6321, build id: 275B1BB4... | 2021-03-24T08:13:23 | 2021-03-24T08:13:23 | {} | MEMBER |
ClickHouse | ClickHouse | 805,794,441 | 20,497 | vdimir | @kolechenkov
Workaround is to disable optimizing predicate expression `SET enable_optimize_predicate_expression = 0`.
We are working on fix in https://github.com/ClickHouse/ClickHouse/pull/20622 I hope it will be merged soon. | 2021-03-24T12:51:42 | 2021-03-24T12:51:42 | {} | MEMBER |
ClickHouse | ClickHouse | 805,819,400 | 22,021 | den-crane | скорее всего 0 не работает как открытый диапазон, что вообще-то логично, с чего бы. | 2021-03-24T13:27:07 | 2021-03-24T13:27:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,921,600 | 21,948 | azat | @kitaisreal instead of simply reverting that patch let's move this under parameter?
Since it really improve performance in some cases.
Initially it was done under parameter, but later @KochetovNicolai was noted that it is useful "almost always".
P.S. Can you please Cc me next if some of my patches will be reverted | 2021-03-24T15:28:03 | 2021-03-24T15:28:03 | {} | MEMBER |
ClickHouse | ClickHouse | 805,963,050 | 17,765 | inakisoriamrf | Hi @ukutaht ,
we use different approaches.
In the row data, we store event_time, event_time_client_locale, client_timezone, the locale time is calculated in the ingestion script.
This allows us to materialize information in one table using event_time_client_locale, clickhouse will think that everything is UTC but yo... | 2021-03-24T16:16:56 | 2021-03-24T16:16:56 | {} | NONE |
ClickHouse | ClickHouse | 806,012,751 | 22,086 | akuzm | >AFAIK deadlocks are possible on Read Committed isolation level when we try to avoid serialization errors by waiting for conflicting transaction to be committed and taking new snapshot for out statement. Maybe we will implement it someday, but it seems to be too hard for now (we still have issues with deadlock detector... | 2021-03-24T17:22:30 | 2021-03-24T17:22:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,145,412 | 22,098 | alexey-milovidov | This will be addressed with lag/lead functions from this PR: #21895 | 2021-03-24T19:58:27 | 2021-03-24T19:58:27 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 806,146,642 | 22,098 | den-crane | Now runningDifference works in multi-threading with unordered blocks. If it will been fixed then it became the same as window function.
```
set max_block_size=3
select number - x from (
SELECT number , any(number) over (rows between 1 preceding and current row) x
FROM numbers(10))
┌─minus(number, x)─┐
│ ... | 2021-03-24T19:59:21 | 2021-03-24T19:59:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,152,459 | 22,098 | den-crane | oops,
it should be between 1 preceding and 1 preceding
```
select number - x from (
SELECT number , any(number) over (rows between 1 preceding and 1 preceding) x
FROM numbers(10))
``` | 2021-03-24T20:05:40 | 2021-03-24T20:05:40 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,163,600 | 18,742 | alexey-milovidov | Although testing with fake time is very valuable for detection of time bombs. | 2021-03-24T20:25:26 | 2021-03-24T20:25:26 | {} | MEMBER |
ClickHouse | ClickHouse | 806,205,549 | 21,600 | alexey-milovidov | I cannot find any traces of what `read_pollfd` is:
```
milovidov@milovidov-desktop:~/work/ClickHouse$ rg read_pollfd
milovidov@milovidov-desktop:~/work/ClickHouse$
```
Maybe it means just the `pollfd` argument of the `poll` function.
And this:
```
Poco::File::exists() const obj-x86_64-linux-gnu/../contrib/p... | 2021-03-24T21:44:02 | 2021-03-24T21:44:02 | {} | MEMBER |
ClickHouse | ClickHouse | 805,386,595 | 21,845 | alexey-milovidov | @CurtizJ I think it's reasonable to implement. | 2021-03-24T00:49:35 | 2021-03-24T00:49:35 | {} | MEMBER |
ClickHouse | ClickHouse | 805,738,015 | 22,069 | filimonov | > Memory limit (total) exceeded
It seems your server has not enough free RAM to run that mutation.
1) How much free RAM do you have when you start the mutation?
2) please share also
```
SELECT * FROM system.mutations
```
| 2021-03-24T11:15:36 | 2021-03-24T11:15:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,884,824 | 22,094 | qoega | similar to #20161? | 2021-03-24T14:50:32 | 2021-03-24T14:50:32 | {} | MEMBER |
ClickHouse | ClickHouse | 805,941,041 | 17,765 | ukutaht | I have a similar problem setting up materialized views in the first place. They work great with no concern for timezones but doing daily aggregates in the customer's configured time zone seems very difficult for me.
@inakivb in your proposed solution, how would you handle the customer changing their reporting timezo... | 2021-03-24T15:50:24 | 2021-03-24T15:50:24 | {} | NONE |
ClickHouse | ClickHouse | 806,157,993 | 9,959 | alexey-milovidov | The function has been added under `timezoneOf` name. | 2021-03-24T20:15:48 | 2021-03-24T20:15:56 | {} | MEMBER |
ClickHouse | ClickHouse | 806,262,505 | 21,943 | alexey-milovidov | I'm also in doubt - isn't it an overkill? | 2021-03-25T00:07:51 | 2021-03-25T00:07:51 | {} | MEMBER |
ClickHouse | ClickHouse | 806,332,140 | 10,943 | marising | hi @yfinkelstein
I take the liberty to ask, how is the feature you describe now progressing? Because we are also troubled by the problem of scaling. | 2021-03-25T03:14:29 | 2021-03-25T03:14:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,938,650 | 21,948 | kitaisreal | @azat filling dictionary source is not first priority case, it seems that we cannot preallocate without risk of preallocating too much size because of non unique elements.
If you can implement preallocation without breaking IDictionarySource abstraction, we can put it under some option. | 2021-03-24T15:47:54 | 2021-03-24T15:47:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,950,780 | 21,599 | filimonov | Too many changes at a time, hard to review / follow the diff.
Some changes done 'just in case', but may have a negative impact.
I would prefer to split that PR into 2 (or more) small parts with clear diff & simple motivation each.
I can do that split on my own - i even almost did it on the top of 46dda06, bu... | 2021-03-24T16:02:02 | 2021-03-24T16:02:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,396,408 | 10,591 | xiaogaozi | I have same question as @vzakaznikov asks. | 2021-03-25T06:23:23 | 2021-03-25T06:23:23 | {} | NONE |
ClickHouse | ClickHouse | 806,437,183 | 19,545 | spff | Same question as above, I would appreciate it if there's any doc mentioning the config reload process (without zookeeper) | 2021-03-25T07:42:34 | 2021-03-25T07:42:34 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 806,488,453 | 22,097 | benbiti | ix broken links in settings.md and rebase master | 2021-03-25T09:12:58 | 2021-03-25T09:13:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,349,548 | 21,907 | alexey-milovidov | I confirm - the code around limiting the chunk size is incorrect.
HTTPServerRequest.cpp:40
> stream = std::make_unique<HTTPChunkedReadBuffer>(std::move(in), max_query_size);
`max_query_size` should not apply to the size of INSERTed data. | 2021-03-23T23:33:20 | 2021-03-23T23:33:20 | {} | MEMBER |
ClickHouse | ClickHouse | 805,461,071 | 21,834 | qoega | I thought about similar things but met several obstacles and left it as it is.
1) we can transform data after generation in select query. This is the way how I used to generate smaller time ranges to allow insert in partitioned tables. Otherwise we will have more parts in one insert.
2) currently data generation unifo... | 2021-03-24T03:34:03 | 2021-03-24T03:34:03 | {} | MEMBER |
ClickHouse | ClickHouse | 805,682,708 | 22,078 | georgthegreat | This patch is a pre-requisite for
https://a.yandex-team.ru/review/1702428/details | 2021-03-24T10:18:26 | 2021-03-24T10:18:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,855,574 | 21,972 | kssenii | @qoega, ok. | 2021-03-24T14:12:53 | 2021-03-24T14:12:53 | {} | MEMBER |
ClickHouse | ClickHouse | 805,993,430 | 21,972 | alexey-milovidov | @qoega It's not possible because third-party ODBC drivers are not instrumented. | 2021-03-24T16:55:48 | 2021-03-24T16:55:48 | {} | MEMBER |
ClickHouse | ClickHouse | 805,546,789 | 12,195 | ucasfl | May be this option should be false directly for `remote` and `cluster` table functions? @alexey-milovidov | 2021-03-24T06:45:44 | 2021-03-24T06:45:44 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 805,823,988 | 22,036 | CurtizJ | Test failures are not related to changes. | 2021-03-24T13:31:03 | 2021-03-24T13:31:03 | {} | MEMBER |
ClickHouse | ClickHouse | 805,854,291 | 21,972 | qoega | Can you check if we can enable MSAN tests in odbc bridge? | 2021-03-24T14:11:17 | 2021-03-24T14:11:17 | {} | MEMBER |
ClickHouse | ClickHouse | 806,191,864 | 22,045 | abyss7 | You should add an integration test for the new feature. | 2021-03-24T21:16:46 | 2021-03-24T21:16:46 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,399,259 | 14,220 | xiaogaozi | Does this PR conflict with #10128? Because #10128 has a task:
> OPTIMIZE FINAL should do everything and get table to the correct state in regards to TTL | 2021-03-25T06:30:03 | 2021-03-25T06:30:03 | {} | NONE |
ClickHouse | ClickHouse | 805,354,397 | 21,902 | alexey-milovidov | This is neither an error nor a bug.
Merges are cancelled in the following cases:
- table is being dropped or detached;
- server shutdown;
- SYSTEM STOP MERGES query is issued;
- (in older versions) when ALTER TABLE is being processed. | 2021-03-23T23:46:12 | 2021-03-23T23:46:12 | {} | MEMBER |
ClickHouse | ClickHouse | 805,538,641 | 22,022 | magvan | > @kitaisreal Наверное Nullable уже работает корректно для `range_hashed` и задачу можно закрывать?
Nullable нет в range_hashed.
Настраиваю словарь:
```
<range_min>
<name>date_start</name>
<type>Date</type>
</range_min>
<range_max>
<name>date_end</name>
<type>... | 2021-03-24T06:28:25 | 2021-03-24T06:28:25 | {} | NONE |
ClickHouse | ClickHouse | 805,541,781 | 22,022 | magvan | > @filimonov В типе данных Date невозможно записать Null поэтому используется значение по умолчанию (1 января 1970).
> Для хранения null в КХ нужно использовать Nullable(...)
Nullable можно использовать в атрибутах. При конфигурировании словаря испльзощвать nullable для полей range_min/range_max нельзя.
Понимаю ... | 2021-03-24T06:35:27 | 2021-03-24T06:35:27 | {} | NONE |
ClickHouse | ClickHouse | 805,544,996 | 22,022 | magvan | Если бы можно было бы использовать **\<expression /\>** в range_min/range_max. Думаю в нем можно было бы заменить null на 1900-01-01. | 2021-03-24T06:42:11 | 2021-03-24T06:42:11 | {} | NONE |
ClickHouse | ClickHouse | 805,744,627 | 22,069 | Air-zone | > > Memory limit (total) exceeded
>
> It seems your server has not enough free RAM to run that mutation.
>
> 1. How much free RAM do you have when you start the mutation?
> 2. please share also
>
> ```
> SELECT * FROM system.mutations
> ```
1. It's a test virtual machine. So there is approximately 3 GB o... | 2021-03-24T11:26:22 | 2021-03-24T11:26:22 | {} | NONE |
ClickHouse | ClickHouse | 805,763,528 | 22,080 | Aureliano-Berlindia | attch the error log:
```
2021.03.24 19:57:17.037580 [ 973 ] {88ff8d8f-02df-443c-9ce4-cd5bb46786a6} <Error> TCPHandler: Code: 1000, e.displayText() = DB::Exception: I/O error: 19, Stack trace:
0. Poco::Net::SocketImpl::error(int, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> ... | 2021-03-24T11:59:03 | 2021-03-24T11:59:03 | {} | NONE |
ClickHouse | ClickHouse | 804,923,399 | 21,904 | godliness | @afiskon
```
git clone git@github.com:ClickHouse/ClickHouse.git
cd ClickHouse
git revert 0832da77eb0a2d01b55cca061a387bf4c833568b
git revert 797a4052d88d6afad832fdfab1f35bbfe29c2244
git submodule update --init --recursive
mkdir build
cd build
cmake .. -DCMAKE_C_COMPILER=`brew --prefix llvm`/bin/clang-11 -DC... | 2021-03-23T13:56:29 | 2021-03-24T01:22:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,546,979 | 21,836 | Jokser | @Avogar Conflicts have resolved. Could you please merge the change? | 2021-03-24T06:46:07 | 2021-03-24T06:46:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 805,886,856 | 21,889 | qoega | AST fuzzer (UBSan) #22093 | 2021-03-24T14:53:04 | 2021-03-24T14:53:04 | {} | MEMBER |
ClickHouse | ClickHouse | 805,934,351 | 21,830 | foolchi | @CurtizJ
test case added, please review again | 2021-03-24T15:42:58 | 2021-03-24T15:42:58 | {} | NONE |
ClickHouse | ClickHouse | 806,147,332 | 22,098 | alexey-milovidov | @den-crane Never thought about this usage of window functions. Thank you! | 2021-03-24T20:00:34 | 2021-03-24T20:00:34 | {} | MEMBER |
ClickHouse | ClickHouse | 806,203,091 | 22,068 | filimonov | Can you compare stacktraces (or perf top / perf record) of your benchmark with and without zookeeper? | 2021-03-24T21:38:53 | 2021-03-24T21:38:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 806,325,791 | 22,054 | amosbird | > There are some caveats:
> - everything after `FORMAT` section is parsed as table data.
`LOAD FILE <filename> INTO <table> [FORMAT ...] `
This statement will be rewritten as `INSERT INTO <table> ...` and have input stream attached to the local file.
| 2021-03-25T02:56:07 | 2021-03-25T02:56:07 | {
"+1": 1
} | COLLABORATOR |
ClickHouse | ClickHouse | 833,535,448 | 23,914 | den-crane | >Sometimes I don't need this merge, and it causes the disks to explode and crash.
CH MergeScheduler is hardcoded and does not have such parameters. Try to reduce background_pool_size (users / default profile), and set it to 8
```
background_pool_size arg Number of threads per... | 2021-05-06T13:45:22 | 2021-05-06T13:45:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 833,550,029 | 23,911 | excitoon | @ucasFL which build of ClickHouse have you tried to reproduce it? | 2021-05-06T14:04:10 | 2021-05-06T14:04:10 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 833,604,199 | 23,196 | kssenii | > Btw why is your progress bar pink?
It is just my terminal color scheme, it overwrites all colors into colors that I chose. | 2021-05-06T15:14:22 | 2021-05-06T15:16:06 | {} | MEMBER |
ClickHouse | ClickHouse | 833,839,277 | 23,921 | kitaisreal | No bug exists in 21.3. | 2021-05-06T20:30:22 | 2021-05-06T20:30:22 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.