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 | 502,120,820 | 5,370 | alesapin | Fixed for stable and prestable releases since v19.7.5.27-stable. | 2019-06-14T14:01:34 | 2019-06-14T14:01:34 | {} | MEMBER |
ClickHouse | ClickHouse | 502,143,935 | 5,618 | den-crane | set in the session using [ set max_bytes_before_external_group_by = .... ]
or users' profiles.
```
cat /etc/clickhouse-server/conf.d/memory_usage.xml
<?xml version="1.0" ?>
<yandex>
<profiles>
<default>
<max_memory_usage_for_all_queries>101310968832</max_memory_usage_for_all_queries>
... | 2019-06-14T15:03:04 | 2019-06-14T15:03:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,197,196 | 5,631 | alexey-milovidov | `clickhouse-benchmark <<< "SELECT count() FROM test.simple_mergetree"`
`sudo perf top --call-graph dwarf`
```
- 61.68% 61.68% libc-2.27.so [.] __memmove_sse2_unaligned_erms ... | 2019-06-14T17:29:48 | 2019-06-14T17:29:48 | {} | MEMBER |
ClickHouse | ClickHouse | 502,230,391 | 5,632 | alexey-milovidov | #5633 (with perf test) | 2019-06-14T19:13:19 | 2019-06-14T19:13:19 | {} | MEMBER |
ClickHouse | ClickHouse | 502,871,787 | 5,661 | alexey-milovidov | We can use more efficient architecture dependent CRC32 implementation, see examples in zlib-ng library, especially "PCLMULQDQ" variant. You can also add a performance test, look at `tests/performance` | 2019-06-17T22:18:03 | 2019-06-17T22:18:03 | {} | MEMBER |
ClickHouse | ClickHouse | 502,944,971 | 5,648 | filimonov | On s3 + fuse (goofys 0.20) performance is up to 3 times better than w/o that change.
Some samples:
```
select count() from tripdata
Before: 1 rows in set. Elapsed: 13.919 sec. Processed 1.31 billion rows, 1.31 GB (93.98 million rows/s., 93.98 MB/s.)
After: 1 rows in set. Elapsed: 6.330 sec. Processed 1.31 bi... | 2019-06-18T04:46:20 | 2019-06-18T04:46:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,072,360 | 5,620 | 4ertus2 | ```
SELECT number FROM system.numbers ANY INNER JOIN (SELECT number, ['test'] FROM system.numbers LIMIT 1) USING (number) LIMIT 1
-- OK
SELECT number FROM system.numbers ANY LEFT JOIN (SELECT number, ['test'] FROM system.numbers LIMIT 1) USING (number) LIMIT 1
Code: 43, e.displayText() = DB::Exception: Neste... | 2019-06-14T11:24:19 | 2019-06-14T11:24:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,127,060 | 5,540 | alexey-milovidov | But it doesn't work for `clickhouse-local`. The only thing I needed:
```
$ clickhouse-local --query "SELECT count() FROM remote('127.0.0.1', system.one)"
Not found: tcp_port
$ clickhouse-local --query "SELECT count() FROM remote('127.0.0.1', system.one)" -- --tcp_port=9000
Not found: tcp_port
``` | 2019-06-14T14:18:36 | 2019-06-14T14:18:49 | {} | MEMBER |
ClickHouse | ClickHouse | 502,483,678 | 5,648 | alexey-milovidov | #5619 #5581 | 2019-06-16T20:33:47 | 2019-06-16T20:33:47 | {} | MEMBER |
ClickHouse | ClickHouse | 502,517,580 | 5,612 | CAspeling | ClickHouse 19.5.3.8
```
select
id, maxIf(value, id == 1) as max
from (
select
number as id,
number * 2 as value
from
numbers(5)
) group by id
```
| id | max |
| 0 | 0 |
| 4 | 0 |
| 3 | 0 |
| 2 | 0 |
| 1 | 2 |
versus
```
select
id, avgIf(value, id == 1) as max
from (
se... | 2019-06-17T02:56:40 | 2019-06-17T03:25:07 | {} | NONE |
ClickHouse | ClickHouse | 502,518,793 | 5,612 | CAspeling | ```
select
id, avgIf(value, id == 1) as max
from (
select
number as id,
CAST(number*2, 'Nullable(Float64)') as value
from
numbers(5)
) group by id
```
| id | max
| 0 |
| 4 |
| 3 |
| 2 |
| 1 | 2
versus
```
select
id,maxIf(value, id == 1) as max
from (
select
number as i... | 2019-06-17T03:05:23 | 2019-06-17T03:25:25 | {} | NONE |
ClickHouse | ClickHouse | 502,201,797 | 5,631 | alexey-milovidov | https://github.com/yandex/ClickHouse/blob/master/CHANGELOG.md#experimental-features
Performance is Ok in 19.5.4 but there is a degradation in later releases. | 2019-06-14T17:43:48 | 2019-06-14T17:43:48 | {} | MEMBER |
ClickHouse | ClickHouse | 502,361,810 | 5,639 | alexey-milovidov | > Removed strange quoting style like `this'.
It is somewhat related to emacs, but the only developer who used emacs have quit. | 2019-06-15T12:13:44 | 2019-06-15T12:13:44 | {} | MEMBER |
ClickHouse | ClickHouse | 502,393,737 | 5,487 | BayoNet | Тут вроде всё хорошо, вмёржить бы :-). | 2019-06-15T19:35:24 | 2019-06-15T19:35:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,522,876 | 5,612 | CAspeling | ```
select id, avgIf(value, id == 1) as avg, maxIf(value, id == 1) as max
from (
select
number as id,
CAST(null, 'Nullable(Float64)') as value
from
numbers(5)
) group by id
```
| id | avg | max |
| 0 | | |
| 4 | | |
| 3 | | |
| 2 | | |
| 1 | | |
and
```
select id, avgIf(val... | 2019-06-17T03:34:21 | 2019-06-17T03:34:21 | {} | NONE |
ClickHouse | ClickHouse | 502,623,735 | 5,645 | 4ertus2 | It's even more interesting. First, we crop real value in phase of parsing
```
select 9223372036854775807.0
SELECT 9223372036854776000.
┌─9223372036854776000.─┐
│ 9223372036854776000 │
└──────────────────────┘
```
Second, even if value could be parsed it will get unexpected conversion error
```
SELECT 922... | 2019-06-17T10:16:41 | 2019-06-17T10:17:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,659,013 | 5,581 | filimonov | Fixed in #5648. Thank you! | 2019-06-17T12:18:39 | 2019-06-17T12:18:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,125,089 | 5,556 | blinkov | @BayoNet в патче какие-то лишние изменения | 2019-06-14T14:13:20 | 2019-06-14T14:13:20 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,141,729 | 5,545 | vavrusa | It looks like that's the intended behavior https://github.com/yandex/ClickHouse/commit/89ffb577b1a1cff5af8b98f48d3a44b810ee0807 and the documentation should not say it's like a boolean.
> There, the exception is caught and logged and since the Stream is not canceled it won't reschedule itself in order to reconnect (... | 2019-06-14T14:57:22 | 2019-06-14T14:57:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,197,988 | 5,631 | alexey-milovidov | `clickhouse-benchmark --max_block_size 1048576 <<< "SELECT count() FROM test.simple_mergetree"`
Increasing `max_block_size` "fixes" this issue. | 2019-06-14T17:32:07 | 2019-06-14T17:35:40 | {} | MEMBER |
ClickHouse | ClickHouse | 502,199,939 | 5,631 | alexey-milovidov | ```
#0 __memmove_sse2_unaligned_erms () at ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:249
#1 0x0000000006ddad87 in std::__copy_move<false, true, std::random_access_iterator_tag>::__copy_m<unsigned long> (__result=<optimized out>, __last=<optimized out>, __first=<optimized out>) at /usr/include/c++/8/b... | 2019-06-14T17:38:13 | 2019-06-14T17:38:24 | {} | MEMBER |
ClickHouse | ClickHouse | 502,648,321 | 1,929 | artemkamolodets | If you have a similar problem (Cannot allocate memory messages in a log files, while there are enough free RAM in the system).
We are running clickhouse in a docker container.
Try to increase the size of /dev/shm partition in the container (shm_size parameter for docker compose) it can helps. | 2019-06-17T11:42:56 | 2019-06-17T11:42:56 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 502,077,074 | 5,621 | filimonov | Add them into user profile in users.xml file. The simplest option - to add them to the `<default>` profile. | 2019-06-14T11:42:26 | 2019-06-14T11:43:14 | {
"-1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,127,437 | 5,556 | BayoNet | Чрезвычайно информативный комментарий, позволяющий сразу понять суть проблемы, локализовать её и решить.
1. В каком из двух патчей?
2. Какие лишние? | 2019-06-14T14:19:39 | 2019-06-14T14:19:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,328,118 | 5,637 | inv2004 | https://github.com/edenhill/librdkafka/pull/2183 | 2019-06-15T02:41:25 | 2019-06-15T02:41:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,393,917 | 5,272 | BayoNet | Тут вроде всё хорошо. Вмёржить бы :-). | 2019-06-15T19:37:54 | 2019-06-15T19:37:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,637,600 | 5,621 | filimonov | Please don't create a separate issue for each tiny question.
It is a continuation of #5618, so let's keep that discussion there. | 2019-06-17T11:05:00 | 2019-06-17T11:05:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,924,057 | 5,612 | CAspeling | That second query, the seo column should have the value 1 on ID 0. This is what we'd expect. That is what I get when I run that query in tabix. We would expect maximum to give nan instead of 0 for all the other IDs but instead it appears that extra maths is necessary as shown in column realmax3 to make this happen whic... | 2019-06-18T02:47:45 | 2019-06-18T02:47:45 | {} | NONE |
ClickHouse | ClickHouse | 502,937,582 | 5,612 | CAspeling | @den-crane exactly. We cannot be dividing by 0. Check my query again. What you have is not what I have.
Something other than 0 divided by something other than 0 will always be 1.
The image I've shown in previous comment shows how it should equal 1. That is only to be used to multiply with maximum to get rid of f... | 2019-06-18T04:07:05 | 2019-06-18T04:10:25 | {} | NONE |
ClickHouse | ClickHouse | 502,100,456 | 5,624 | alesapin | Diff is large because I split `MergedBlockOutputStream` into 3 different files.
The main idea is to write special **final** mark after end of column and add it to `index_granularity` with zero rows. After that in `MergeTreeDataSelectExecutor` we can check that open range of marks ends with final mark. So we can use... | 2019-06-14T13:01:28 | 2019-06-14T13:01:28 | {} | MEMBER |
ClickHouse | ClickHouse | 502,102,648 | 5,615 | v1rusw0rm | The bug only happens if nothing was selected\fetched from the table. Log after SELECT:
```
2019.06.14 17:04:04.375179 [ 41 ] {9e9eb6a5-94e7-4917-a0fb-5108050e2b4c} <Debug> executeQuery: (from 127.0.0.1:31154) SELECT * FROM kafka_raw6 LIMIT 1;
2019.06.14 17:04:04.375773 [ 41 ] {9e9eb6a5-94e7-4917-a0fb-5108050e2b4c} <... | 2019-06-14T13:08:14 | 2019-06-14T13:08:14 | {} | NONE |
ClickHouse | ClickHouse | 502,128,555 | 5,556 | blinkov | наверное в первом, сейчас уже норм | 2019-06-14T14:22:34 | 2019-06-14T14:22:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,355,584 | 5,605 | prenc | Stupid error in *schema.xml*, I misunderstood clickhouse documentation. **<source_cluster>** tag should be replaced with real name of the cluster (<real_name>) under **<remote_servers>**. | 2019-06-15T10:39:01 | 2019-06-15T10:44:34 | {
"+1": 3
} | NONE |
ClickHouse | ClickHouse | 502,226,181 | 5,630 | alexey-milovidov | TODOs:
(fixed) It doesn't support Date and DateTime:
```
$ curl 'http://localhost:8123/?param_s=2019-01-01+00:00:00' -d "SELECT {s:DateTime('UTC')}"
Code: 36, e.displayText() = DB::Exception: Value 2019-01-01 00:00:00 cannot be parsed as DateTime('UTC') for query parameter 's' (version 19.10.1.1)
```
(fixed... | 2019-06-14T18:59:37 | 2019-06-15T18:57:37 | {} | MEMBER |
ClickHouse | ClickHouse | 502,393,763 | 5,461 | BayoNet | Тут вроде всё хорошо. Вмёржить бы :-). | 2019-06-15T19:35:44 | 2019-06-15T19:35:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,430,788 | 5,607 | coraxster | Got some non-fixed boundary cases. Made a new PR.
https://github.com/yandex/ClickHouse/pull/5645 | 2019-06-16T08:09:18 | 2019-06-16T08:09:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,506,287 | 5,646 | yurriy | > If I remember correctly, `setUser` will be enough?
Also replacing strings by PODArray, which uses MemoryTracker. | 2019-06-17T01:39:15 | 2019-06-17T01:39:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,719,828 | 5,582 | blind-oracle | @Felixoid
The timestamp didn't help either.
When the problem manifests I'm seeing the following strange entries in the Graphite table:
```
│ 1560782520 │ 2019-06-17 16:42:00 │ 64 │
│ 1560782580 │ 2019-06-17 16:43:00 │ 64 │
│ 1560782580 │ 2019-06-17 16:43:00 │ 1168 │
│ 1560782668 │ 2019-06-17 16:44:28... | 2019-06-17T14:59:41 | 2019-06-17T14:59:41 | {} | NONE |
ClickHouse | ClickHouse | 502,938,779 | 5,612 | CAspeling | The seo column is just something we came up with to show what we expect maxIf to return. There is still the problem of avgIf and maxIf behaving differently according to input. It would solve so many of our problems if this was resolved. We wouldn't need to implement a hack like that realmax3 and seo columns just to get... | 2019-06-18T04:13:34 | 2019-06-18T04:13:34 | {} | NONE |
ClickHouse | ClickHouse | 502,073,879 | 5,616 | filimonov | Continuation in #5618 | 2019-06-14T11:30:21 | 2019-06-14T11:30:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,294,593 | 5,605 | filimonov | Can you put a line with comment why did you close the issue? | 2019-06-14T22:05:15 | 2019-06-14T22:05:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,357,213 | 5,633 | alex-zaitsev | Backport to 19.6/7/8? | 2019-06-15T11:05:16 | 2019-06-15T11:05:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,434,525 | 5,484 | BayoNet | Здесь вроде бы всё хорошо. Вмёржить бы :-) | 2019-06-16T09:07:51 | 2019-06-16T09:07:51 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,600,488 | 5,402 | blinkov | @BayoNet docs build still seems broken | 2019-06-17T09:09:30 | 2019-06-17T09:09:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,647,169 | 1,929 | I-akshay | > dividing the original queue to smaller parts
Hi artemkamolodets , can you please explain how to divide the original queue to smaller parts. | 2019-06-17T11:38:30 | 2019-06-17T11:38:30 | {} | NONE |
ClickHouse | ClickHouse | 502,937,156 | 5,612 | den-crane | >That second query, the seo column should have the value 1 on ID 0.
No [ / (average + 1e-14)) AS seo ] === / ( -1e-14 + 1e-14 ) ~ nan of inf
because of math
```
SELECT 1 / 0
┌─divide(1, 0)─┐
│ inf │
└──────────────┘
SELECT 0 / 0
┌─divide(0, 0)─┐
│ nan │
└──────────────┘
``` | 2019-06-18T04:04:15 | 2019-06-18T04:04:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,075,195 | 5,589 | filimonov | a) just create tables clusterwise using ON CLUSTER clause and you will have no such problems
b) you can either insert into distributed table which has an sharding key controlling where to put the data (you can use random() to distrinute data equally). Or you can insert directly to underlying tables (usually mergetree ... | 2019-06-14T11:35:15 | 2019-06-14T11:35:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,131,815 | 5,545 | spyros87 | @CurtizJ I have made an observation regarding the issue. The documentation in EN mentions:
> kafka_skip_broken_messages – Kafka message parser mode. If kafka_skip_broken_messages = 1 then the engine skips the Kafka messages that can't be parsed (a message equals a row of data).
Which I have interpreted as a `bool... | 2019-06-14T14:31:00 | 2019-06-14T14:31:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,372,086 | 5,641 | alexey-milovidov | Yes. | 2019-06-15T14:38:22 | 2019-06-15T14:38:22 | {} | MEMBER |
ClickHouse | ClickHouse | 502,635,757 | 5,618 | filimonov | @I-akshay , yes that file should be created.
@den-crane provided a full content of that file. I don't understand why can you create by your own and need to download it somewhere, but please take it here: [memory_usage.zip](https://github.com/yandex/ClickHouse/files/3296523/memory_usage.zip)
You need to:
1) unpa... | 2019-06-17T10:58:36 | 2019-06-17T11:05:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,457,806 | 5,565 | Slind14 | I see. Is LowCardinality or Null the cause? | 2019-06-16T14:34:28 | 2019-06-16T14:34:28 | {} | NONE |
ClickHouse | ClickHouse | 502,532,980 | 5,648 | filimonov | For benchmarks - try `select * from hugetable limit 1` (the visible delay on ext4 with such selects looks pretty the same as almost any select on fuse).
| 2019-06-17T04:47:24 | 2019-06-17T04:47:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,817,387 | 5,657 | alexey-milovidov | Why it's closed? | 2019-06-17T19:24:05 | 2019-06-17T19:24:05 | {} | MEMBER |
ClickHouse | ClickHouse | 502,458,100 | 5,565 | simPod | For me it was UUID type | 2019-06-16T14:38:04 | 2019-06-16T14:38:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,517,841 | 5,648 | jackpgao | Got, I'll test it today. | 2019-06-17T02:58:21 | 2019-06-17T02:58:21 | {} | NONE |
ClickHouse | ClickHouse | 502,562,383 | 5,648 | filimonov | Same sample table, Ext4, SSD, `select * from bad_partitions limit 1`.
Before:
```
echo 'select * from bad_partitions limit 1' | clickhouse-benchmark -i 20 -d 0
Loaded 1 queries.
Queries executed: 20.
QPS: 0.685, RPS: 0.685, MiB/s: 0.000, result RPS: 0.685, result MiB/s: 0.000.
0.000% 1.427 sec.
10.000% 1.... | 2019-06-17T07:11:52 | 2019-06-17T07:25:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,124,223 | 5,612 | den-crane | What CH version do you use? I don't observe the described behavior.
```
SELECT CAST(NULL, 'Nullable(Float64)') AS a
FROM numbers(3)
┌────a─┐
│ ᴺᵁᴸᴸ │
│ ᴺᵁᴸᴸ │
│ ᴺᵁᴸᴸ │
└──────┘
SELECT max(a)
FROM
(
SELECT CAST(NULL, 'Nullable(Float64)') AS a
FROM numbers(3)
)
┌─max(a)─┐
│ ᴺᵁᴸᴸ │
└────... | 2019-06-14T14:10:59 | 2019-06-14T14:10:59 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,483,618 | 5,648 | alexey-milovidov | But it will definitely help @filimonov to test ClickHouse on S3 over FUSE (that's not a good idea BTW). | 2019-06-16T20:33:04 | 2019-06-16T20:33:04 | {} | MEMBER |
ClickHouse | ClickHouse | 502,524,516 | 5,612 | CAspeling | In fact, ifnull doesn't want to output the expected "1" either but the following at least gives the last column for max to be nulls
```
select
id,
avgIf(value, id == 3) as average,
maxIf(value, id == 3) as maximum,
ifNull(average,1) as thing,
maximum*thing/thing as realmax
from (
select
number a... | 2019-06-17T03:45:51 | 2019-06-17T04:30:18 | {} | NONE |
ClickHouse | ClickHouse | 502,933,912 | 5,612 | CAspeling | > So the difference is that avg(0/0) == null ? It seems already fixed.
Can't divide by 0, hence the maths to prevent that. | 2019-06-18T03:44:18 | 2019-06-18T03:44:18 | {} | NONE |
ClickHouse | ClickHouse | 502,372,054 | 5,641 | blinkov | @alexey-milovidov shouldn't this update documentation as well? | 2019-06-15T14:37:56 | 2019-06-15T14:37:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,393,672 | 5,566 | BayoNet | Тут вроде всё хорошо. Вмёржить бы :-). | 2019-06-15T19:34:30 | 2019-06-15T19:34:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,393,783 | 5,459 | BayoNet | Тут вроде всё хорошо. Вмёржить бы :-). | 2019-06-15T19:36:04 | 2019-06-15T19:36:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,483,488 | 5,648 | alexey-milovidov | I don't see much difference on my tests: `stat` syscalls on ext4 are very cheap.
```
CREATE TABLE bad_partitions (a UInt64, b UInt64, c UInt64, d UInt64, e UInt64, f UInt64, g UInt64, h UInt64, i UInt64, j UInt64, k UInt64, l UInt64, m UInt64, n UInt64, o UInt64, p UInt64, q UInt64, r UInt64, s UInt64, t UInt64, u ... | 2019-06-16T20:30:56 | 2019-06-16T20:30:56 | {} | MEMBER |
ClickHouse | ClickHouse | 502,075,753 | 5,619 | filimonov | Probably that is more important: #5581 | 2019-06-14T11:37:24 | 2019-06-14T11:37:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,134,068 | 5,450 | alexey-milovidov | @geldot Not quite. Because "external" data can be read multiple times in a query and must be buffered. Typical example is `SELECT ... WHERE x GLOBAL IN _table` - a table is send to multiple shards.
With `input` table function it is specifically prohibited. And `input` looks very naturally as the data that client sen... | 2019-06-14T14:37:20 | 2019-06-14T14:37:20 | {} | MEMBER |
ClickHouse | ClickHouse | 502,139,069 | 5,618 | I-akshay | > Alternative option - is to do huge group by / order by operations on disks. It is slower, but allow to execute those operations if the data doesn't fit in memory.
How to do this ?? | 2019-06-14T14:50:32 | 2019-06-14T14:50:32 | {} | NONE |
ClickHouse | ClickHouse | 502,218,046 | 5,631 | alexey-milovidov | 19.5.4:
```
1 rows in set. Elapsed: 0.423 sec. Processed 10.00 billion rows, 20.00 GB (23.64 billion rows/s., 47.29 GB/s.)
```
This is the right performance.
19.9:
```
1 rows in set. Elapsed: 11.018 sec. Processed 10.00 billion rows, 20.00 GB (907.60 million rows/s., 1.82 GB/s.)
```
This is more tha... | 2019-06-14T18:33:51 | 2019-06-14T18:34:13 | {
"heart": 1
} | MEMBER |
ClickHouse | ClickHouse | 502,407,245 | 5,525 | alexey-milovidov | Follow-up: #5643 | 2019-06-15T23:38:02 | 2019-06-15T23:38:02 | {} | MEMBER |
ClickHouse | ClickHouse | 502,480,319 | 5,648 | blinkov | @jackpgao could you test this patch to see if it helps with partition deletion issues you were talking about? | 2019-06-16T19:46:26 | 2019-06-16T19:46:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,633,721 | 5,650 | filimonov | Please don't create a separate issue for each tiny question.
It is a continuation of #5618, so let's keep that discussion there. | 2019-06-17T10:51:16 | 2019-06-17T10:51:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,073,622 | 5,618 | filimonov | 1) OOMKiller killed the clickhouse process (check dmesg in the shell).
If those parameters are zero (default) - clickhouse tries do to all group by & order by in memory (it will allocate up to `max_memory_usage` bytes, you probably have less than 10Gb of RAM).
Alternative option - is to do huge group by / order ... | 2019-06-14T11:29:13 | 2019-06-14T11:29:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,093,785 | 5,519 | Chocorean | Well, I did not manage to recreate such a situation after changing my partition key (from `domain`, around 1m5 partitions to `date`, 1 per day).
It is good I have no more segfault but I probably should not crash when having a lot of partitions.
Feel free to close this if you won't investigate further, I no longer... | 2019-06-14T12:39:52 | 2019-06-14T12:39:52 | {} | NONE |
ClickHouse | ClickHouse | 502,103,300 | 5,604 | simPod | Understand, thanks. Maybe the message could be more explicit? | 2019-06-14T13:10:08 | 2019-06-14T13:10:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,152,912 | 4,631 | vavrusa | > We are probably facing same issue, I have set kafka_num_consumers to match number of partitions(2)
You should always set the `kafka_num_consumers` to be lower than the `number_of_kafka_partitions / number_of_clickhouse_nodes`. If your Kafka topic has 10 partitions for example and you have 2 clickhouse nodes, you s... | 2019-06-14T15:26:45 | 2019-06-14T15:26:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,343,775 | 5,519 | Chocorean | 1 million 500 thousands partitions* with `domain` as partition key | 2019-06-15T07:41:45 | 2019-06-15T07:41:45 | {} | NONE |
ClickHouse | ClickHouse | 502,431,103 | 5,645 | coraxster | @4ertus2 could you please to take a look? | 2019-06-16T08:13:38 | 2019-06-16T08:13:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,554,660 | 5,532 | blinkov | Yes, fixed a while ago, but forgot to close in time. | 2019-06-17T06:42:15 | 2019-06-17T06:42:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,579,361 | 5,582 | blind-oracle | @Felixoid
Thanks, in my case `carbon-clickhouse` uploads zero there, so I'll reconfigure it to write a timestamp.
Aside from this, I've turned off replication (used `GraphiteMergeTree` instead of `Replicated...`). The averaging errors almost went away and are now *in sync* between two clusters.
Cluster1:
```
... | 2019-06-17T08:09:02 | 2019-06-17T08:09:02 | {} | NONE |
ClickHouse | ClickHouse | 502,633,172 | 5,649 | filimonov | Please don't create a separate issue for each tiny question.
It is a continuation of #5618, let's keep that discussion there. | 2019-06-17T10:49:23 | 2019-06-17T10:50:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,667,742 | 5,636 | stavrolia | @alexey-milovidov, was there any .xml file in your ~/ClickHouse/dbms/tests/tmp directory? | 2019-06-17T12:46:22 | 2019-06-17T12:46:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,684,557 | 5,635 | stavrolia | @alexey-milovidov There is an option --input-files to set the list of input-files. clickhouse-performance-test --help shows usage like "clickhouse-performance-test [options]" and it works. Is this issue about changing current api or name of option? | 2019-06-17T13:34:50 | 2019-06-17T13:34:50 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,689,266 | 5,612 | den-crane | So the difference is that avg(0/0) == null ? It seems already fixed.
Current behavior is nan for avg in case of 0 count.
19.10.1.669
```
SELECT
id,
avgIf(value, id = 1) AS max,
sumIf(value, id = 1) AS sum,
countIf(value, id = 1) AS cnt
FROM
(
SELECT
number AS id,
num... | 2019-06-17T13:47:37 | 2019-06-17T13:50:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,071,223 | 5,616 | filimonov | OOMKiller killed the clickhouse process (check `dmesg` in the shell). | 2019-06-14T11:19:54 | 2019-06-14T11:19:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,138,795 | 5,621 | I-akshay | > Add them into user profile in users.xml file. The simplest option - to add them to the `<default>` profile.
I am currently using default profile | 2019-06-14T14:49:49 | 2019-06-14T14:49:49 | {} | NONE |
ClickHouse | ClickHouse | 502,383,457 | 5,630 | alexey-milovidov | TODO: (postponed) it's not supported for clickhouse-local. | 2019-06-15T17:09:52 | 2019-06-15T18:57:48 | {} | MEMBER |
ClickHouse | ClickHouse | 502,393,718 | 5,563 | BayoNet | Тут вроде всё хорошо. Вмёржить бы. | 2019-06-15T19:35:04 | 2019-06-15T19:35:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,537,581 | 5,618 | I-akshay | > set in the session using [ set max_bytes_before_external_group_by = .... ]
> or users' profiles.
>
> ```
> cat /etc/clickhouse-server/conf.d/memory_usage.xml
> <?xml version="1.0" ?>
> <yandex>
> <profiles>
> <default>
> <max_memory_usage_for_all_queries>101310968832</max_memory_usag... | 2019-06-17T05:17:17 | 2019-06-17T05:25:06 | {} | NONE |
ClickHouse | ClickHouse | 502,550,840 | 5,532 | andyyzh | it is ready. | 2019-06-17T06:25:55 | 2019-06-17T06:25:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,751,910 | 5,582 | den-crane | Try to add MV with plain MergeTree (from shard_02.graphite) and validate that these duplicates are not from carbon-clickhouse | 2019-06-17T16:18:08 | 2019-06-17T16:18:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,817,640 | 5,657 | alexey-milovidov | Yes, we need to rebase it on master. | 2019-06-17T19:24:56 | 2019-06-17T19:24:56 | {} | MEMBER |
ClickHouse | ClickHouse | 502,829,770 | 5,582 | blind-oracle | @den-crane
Will try, but I'm 99.9% sure they aren't.
Because:
* The metrics are sent in specific seconds (like 28/58), not in 0th second.
* Neighboring MergeTree table gets the same data and it does not have duplicates | 2019-06-17T20:02:31 | 2019-06-17T20:02:31 | {} | NONE |
ClickHouse | ClickHouse | 502,534,797 | 5,612 | CAspeling | The following maths work because it uses a number so small that the db can't actually ever store or use it so we never run into the problem of /0
```
select
id,
value,
avgIf(value, id == 0) as average,
maxIf(value, id == 0) as maximum,
( ( (average-0.00000000000001) / (average+0.00000000000001) ) / (... | 2019-06-17T04:59:15 | 2019-06-18T04:07:55 | {} | NONE |
ClickHouse | ClickHouse | 502,104,141 | 5,519 | den-crane | What is 1m5 stands for? | 2019-06-14T13:12:52 | 2019-06-14T13:12:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,445,945 | 5,600 | alexey-milovidov | https://clickhouse-test-reports.s3.yandex.net/5600/d1904c251ee29547eb12bb2e5b7a85379b992739/pvs_studio_report/pvs-studio-html-report/sources/BitHelpers.cpp_32.html#ln102 | 2019-06-16T12:01:35 | 2019-06-16T12:01:35 | {} | MEMBER |
ClickHouse | ClickHouse | 502,464,825 | 5,541 | vzakaznikov | Hi @KochetovNicolai, I have addressed all the failing checks and the PR is ready for your review. The only fail left does not seem to be caused by the changes related to LIVE VIEW tables. Thanks. | 2019-06-16T16:05:20 | 2019-06-16T16:05:20 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,465,904 | 5,646 | yurriy | > Great! PS. Did you check (manually) that the user settings (e. g. `max_memory_usage` specified in `users.xml`) work?
No, the last thing to do from #5476 is implementing user settings. I was going to make them separately. | 2019-06-16T16:20:36 | 2019-06-16T16:22:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 502,472,848 | 5,646 | alexey-milovidov | If I remember correctly, `setUser` will be enough? | 2019-06-16T17:59:20 | 2019-06-16T17:59:20 | {} | MEMBER |
ClickHouse | ClickHouse | 502,639,660 | 5,402 | BayoNet | It's required another pull to be merged to resolve links correctly. After the master branch merge the error is disappeared. | 2019-06-17T11:12:13 | 2019-06-17T11:12:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 530,889,802 | 3,490 | alexey-milovidov | @akuzm Maybe it depends on Linux kernel. Just for the reference, you can tell what version of linux kernel is used on that machine. | 2019-09-12T15:53:45 | 2019-09-12T15:53:45 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.