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 | 656,872,791 | 8,685 | alexey-milovidov | @mnach Yes, it's totally alright.
(Also ClickHouse should continue to work when there is no systemd in the system). | 2020-07-10T20:21:29 | 2020-07-10T20:21:29 | {} | MEMBER |
ClickHouse | ClickHouse | 657,179,868 | 4,415 | alexey-milovidov | The data in table with `Log` engine gets corrupted.
Note that `Log` table engine does not maintain data consistency. This situation is normal and documented:
https://clickhouse.tech/docs/en/engines/table-engines/log-family/
You should use full-featured table engines like ReplicatedMergeTree or MergeTree. | 2020-07-12T06:05:19 | 2020-07-12T06:05:19 | {} | MEMBER |
ClickHouse | ClickHouse | 657,226,037 | 12,429 | den-crane | Zstandard includes Huffman coding. You can use ZSTD CODEC. | 2020-07-12T14:02:45 | 2020-07-12T14:02:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,259,833 | 5,462 | mzealey | In my tests, on disk when we use String to store 4 bytes of data ipv4 ip address, the compressed size is roughly 1 byte (~2.3 bytes) smaller than using `FixedString(16)`. I'm not using the `::ffff` style syntax just storing 4 or 16 bytes (99% ipv4). | 2020-07-12T18:42:07 | 2020-07-12T18:42:07 | {} | NONE |
ClickHouse | ClickHouse | 656,932,729 | 12,415 | den-crane | Materialized view (MV) is a post-insert trigger. Insert to a source table pushes an inserted buffer to MV as well.
MV does select over the inserted buffer (MV never reads the source table except populate stage).
MV does not see `alter update/delete`.
MV does not see changes changes from merge process collapsing... | 2020-07-10T23:09:46 | 2020-07-10T23:10:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 656,936,371 | 12,415 | mauidude | Thanks for answering that, I couldn't find it in the docs anywhere. Is there any way to get atomicity between a table and a materialized view? | 2020-07-10T23:25:28 | 2020-07-10T23:25:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 656,975,504 | 11,815 | bohutang | > It is already mentioned as
>
> > Logical error, can be reproduced with 00993_system_parts_race_condition_drop_zookeeper and https://clickhouse-test-reports.s3.yandex.net/0/90b5e5d2f9edbcd4f01bd596ec919a32368fd716/functional_stateless_tests_(debug).html#fail1
Aha, maybe we should add a failure-collector-robot to... | 2020-07-11T03:05:58 | 2020-07-11T03:05:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,143,767 | 11,541 | bobrik | I've rechecked `v20.6.1.4050-testing` and the issue seems to be gone now 🎉 | 2020-07-11T23:02:43 | 2020-07-11T23:02:43 | {
"hooray": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,165,381 | 10,966 | alexey-milovidov | This library can be way better for level 1: http://www.libslz.org/ | 2020-07-12T02:40:16 | 2020-07-12T02:40:16 | {} | MEMBER |
ClickHouse | ClickHouse | 657,179,577 | 4,373 | alexey-milovidov | I'm sure that this issue does not exist in recent releases. | 2020-07-12T06:01:01 | 2020-07-12T06:01:01 | {} | MEMBER |
ClickHouse | ClickHouse | 657,256,101 | 10,697 | alexey-milovidov | #12437 | 2020-07-12T18:06:44 | 2020-07-12T18:06:44 | {} | MEMBER |
ClickHouse | ClickHouse | 657,259,475 | 12,402 | alexey-milovidov | Let's disable Log, TinyLog, StripeLog. | 2020-07-12T18:38:45 | 2020-07-12T18:38:45 | {} | MEMBER |
ClickHouse | ClickHouse | 657,276,479 | 12,424 | azat | >But this simple test passes:
Yep clickhouse-client everything works, and the query does contain the setting in the query_log.
I think here is the issue, perf tests uses clickhouse-driver in python, which just sends the query and this is not enough for this setting.
While clickhouse-client parse the settings fro... | 2020-07-12T21:17:29 | 2020-07-12T21:17:29 | {} | MEMBER |
ClickHouse | ClickHouse | 656,843,371 | 10,793 | alexey-milovidov | The task is not so easy and no one needs it. | 2020-07-10T19:09:43 | 2020-07-10T19:09:43 | {} | MEMBER |
ClickHouse | ClickHouse | 656,856,416 | 10,824 | alexey-milovidov | @qoega ? | 2020-07-10T19:39:36 | 2020-07-10T19:39:36 | {} | MEMBER |
ClickHouse | ClickHouse | 656,930,314 | 12,409 | den-crane | @adrian17 good catch, I did not know that. | 2020-07-10T22:59:41 | 2020-07-10T22:59:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,179,370 | 4,456 | alexey-milovidov | @alesapin Do I remember correctly that we have observed similar issue recently? | 2020-07-12T05:57:43 | 2020-07-12T05:57:43 | {} | MEMBER |
ClickHouse | ClickHouse | 657,215,739 | 8,834 | amosbird | It seems nullable keys are valid after this pr. I'll help finish what @alexey-milovidov proposed | 2020-07-12T12:31:59 | 2020-07-12T12:31:59 | {
"+1": 1
} | COLLABORATOR |
ClickHouse | ClickHouse | 657,274,503 | 12,424 | azat | >But paradoxically, most of the queries that become faster have fixed block_size specified in SETTINGS clause.
This is interesting.
I got one query from perf tests, replaced hits_10m_single with hits_100m_obfuscated, and got this:
- clickhouse.orig (from upstream/master): with adjusted block-size to take the p... | 2020-07-12T21:01:37 | 2020-07-12T21:01:37 | {} | MEMBER |
ClickHouse | ClickHouse | 656,846,544 | 8,026 | alexey-milovidov | CC @traceon | 2020-07-10T19:14:25 | 2020-07-10T19:14:25 | {} | MEMBER |
ClickHouse | ClickHouse | 656,860,014 | 11,853 | alexey-milovidov | @blinkov You forgot about this PR. | 2020-07-10T19:48:50 | 2020-07-10T19:48:50 | {} | MEMBER |
ClickHouse | ClickHouse | 656,164,688 | 12,323 | alexey-milovidov | Does it reproduce in 20.5, 20.4, 20.3, 19.14? | 2020-07-09T14:34:13 | 2020-07-10T20:12:58 | {} | MEMBER |
ClickHouse | ClickHouse | 656,960,428 | 12,415 | mauidude | Thanks for pointing that out. Ok so if I understand correctly, by enabling that setting, if that scenario happens where an insert succeeds in the table but not the MV, the client would receive an error and would need to retry the insert. When it retries, the table will see it as a duplicate insert and ignore it but the... | 2020-07-11T01:30:08 | 2020-07-11T01:30:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,261,973 | 9,881 | alexey-milovidov | Multiplication of float * decimal should work and produce float.
Reason: multiplication by float implies loss of precision.
Interpretation of literals like `0.1` as float or decimal is completely different task and can be solved separately. | 2020-07-12T19:02:55 | 2020-07-12T19:02:55 | {
"eyes": 1
} | MEMBER |
ClickHouse | ClickHouse | 656,837,279 | 8,101 | alexey-milovidov | It should work with official binary: https://clickhouse.tech/docs/en/getting-started/install/#from-binaries-non-linux
But we have not validated it. It will be appreciated if you will try the official binary and share the findings. | 2020-07-10T19:03:41 | 2020-07-10T19:04:09 | {} | MEMBER |
ClickHouse | ClickHouse | 656,840,332 | 9,461 | alexey-milovidov | You can use the official pre-built binary for FreeBSD:
https://clickhouse.tech/docs/en/getting-started/install/#from-binaries-non-linux
| 2020-07-10T19:06:37 | 2020-07-10T19:06:37 | {} | MEMBER |
ClickHouse | ClickHouse | 656,849,693 | 6,898 | alexey-milovidov | We have replaced LibreSSL with OpenSSL since version 20.1. | 2020-07-10T19:22:13 | 2020-07-10T19:22:13 | {} | MEMBER |
ClickHouse | ClickHouse | 656,850,733 | 11,568 | alexey-milovidov | Could you please run server with `gdb`:
```
gdb --args ./clickhouse server
```
Type `run`.
Wait for the error to happen.
Type `backtrace`.
and
`thread apply all backtrace`
and send the output here? | 2020-07-10T19:24:47 | 2020-07-10T19:24:47 | {} | MEMBER |
ClickHouse | ClickHouse | 656,854,431 | 11,642 | alexey-milovidov | @abyss7 How does it happen? | 2020-07-10T19:34:15 | 2020-07-10T19:34:15 | {} | MEMBER |
ClickHouse | ClickHouse | 657,165,461 | 10,966 | alexey-milovidov | Also worth looking to https://github.com/ebiggers/libdeflate.
It has incompatible with zlib interface, but it's actually better (allows to avoid useless memory copy into sliding window). | 2020-07-12T02:41:25 | 2020-07-12T02:41:25 | {} | MEMBER |
ClickHouse | ClickHouse | 657,175,176 | 12,390 | etah000 | OK, I will add two setting parameters and make the necessary changes.
| 2020-07-12T04:53:30 | 2020-07-12T04:56:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,214,042 | 12,317 | zhbdesign | > Network timeout happens before optimise finishes. Check timeout settings both on server and client
Is there any other way to solve this problem? | 2020-07-12T12:16:08 | 2020-07-12T12:16:08 | {} | NONE |
ClickHouse | ClickHouse | 657,258,377 | 12,278 | alexey-milovidov | > Performance — 1 faster, 7 slower
Now totally different set of tests become slower. | 2020-07-12T18:28:00 | 2020-07-12T18:28:00 | {} | MEMBER |
ClickHouse | ClickHouse | 657,259,815 | 1,825 | alexey-milovidov | Now Ok:
> Outputs the content of the system.processes table, that contains a list of queries that is being processed at the moment, **excepting SHOW PROCESSLIST queries**. | 2020-07-12T18:42:02 | 2020-07-12T18:42:02 | {} | MEMBER |
ClickHouse | ClickHouse | 656,852,120 | 5,675 | alexey-milovidov | CC @qoega @alesapin | 2020-07-10T19:28:20 | 2020-07-10T19:28:20 | {} | MEMBER |
ClickHouse | ClickHouse | 656,860,515 | 10,528 | alexey-milovidov | @vitlibar Outdated? | 2020-07-10T19:50:06 | 2020-07-10T19:50:06 | {} | MEMBER |
ClickHouse | ClickHouse | 656,874,221 | 11,005 | alexey-milovidov | @filimonov Could you please provide an example on how exactly edit the .spec file? | 2020-07-10T20:25:01 | 2020-07-10T20:25:13 | {} | MEMBER |
ClickHouse | ClickHouse | 656,885,249 | 11,642 | abyss7 | I will take a look | 2020-07-10T20:53:06 | 2020-07-10T20:53:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 656,916,384 | 12,409 | den-crane | > (You can work around this by having a tuple (k1, k2) column and using it as a key.)
it's not true.
```
CREATE TABLE jtest
(
`a` String,
`b` String,
`c` Float64
)
ENGINE = Join(any, left, a, b)
insert into jtest values('1','1',0),('2','2',1)
select joinGet('jtest', 'c', tuple('2','2'... | 2020-07-10T22:12:10 | 2020-07-10T22:17:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,113,627 | 12,415 | den-crane | A1: Yes.
A2: Doc: _This behaviour exists to enable insertion of highly aggregated data into materialized views, for cases where inserted blocks are the same after materialized view aggregation but derived from different INSERTs into the source table._
https://github.com/ClickHouse/ClickHouse/issues/5274 | 2020-07-11T19:09:16 | 2020-07-11T19:09:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,156,912 | 9,248 | alexey-milovidov | Everything looks perfect so far.
But there are some inefficiencies:
1. The INSERT SELECT query has squashing of blocks: SELECT produces blocks of 65536 rows and they are squashed to 1048576 rows before INSERT. This can be avoided and PR #12195 will address this.
2. IO buffers of 1 MiB are allocated and freed f... | 2020-07-12T01:02:03 | 2020-07-12T01:02:03 | {} | MEMBER |
ClickHouse | ClickHouse | 657,253,560 | 8,834 | alexey-milovidov | Ok. And please add a setting to alllow/disallow Nullable in primary key, because allowing them contradicts with SQL standard or usual behaviour of other DBMS. | 2020-07-12T17:40:01 | 2020-07-12T17:40:01 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 656,845,263 | 9,902 | alexey-milovidov | After https://github.com/ClickHouse/ClickHouse/pull/12181 we don't need `flatc` for build anymore. | 2020-07-10T19:11:39 | 2020-07-10T19:11:39 | {} | MEMBER |
ClickHouse | ClickHouse | 656,870,264 | 9,925 | alexey-milovidov | The is no global architecture diagram of ClickHouse.
But there are diagrams for various parts:
- indexes and MergeTree;
- distributed query processing;
- replication.
They can be found in presentations published here: https://github.com/ClickHouse/clickhouse-presentations | 2020-07-10T20:15:13 | 2020-07-10T20:15:13 | {} | MEMBER |
ClickHouse | ClickHouse | 656,912,877 | 12,409 | adrian17 | (You can work around this by having a tuple (k1, k2) column and using it as a key.) | 2020-07-10T22:04:09 | 2020-07-10T22:04:09 | {} | NONE |
ClickHouse | ClickHouse | 657,128,652 | 12,420 | alexey-milovidov | CC @MovElb | 2020-07-11T20:57:09 | 2020-07-11T20:57:09 | {} | MEMBER |
ClickHouse | ClickHouse | 657,154,483 | 12,422 | alexey-milovidov | Succeeded:
> Push to dockerhub — Updated yandex/clickhouse-performance-comparison | 2020-07-12T00:37:25 | 2020-07-12T00:37:25 | {} | MEMBER |
ClickHouse | ClickHouse | 657,190,503 | 5,462 | mzealey | Yes, however storing the value as a `String` uses less space than as a `FixedString(16)` if the data is predominantly IPv4. Perhaps could do this and then some sort of detection based on length being 4 or 16 bytes to figure out what the input/output format would be? | 2020-07-12T08:21:45 | 2020-07-12T08:21:45 | {} | NONE |
ClickHouse | ClickHouse | 657,262,058 | 9,881 | alexey-milovidov | When multiplying float by decimal, any kind of decimal should be converted to Float64 before multiplication. | 2020-07-12T19:03:42 | 2020-07-12T19:03:42 | {} | MEMBER |
ClickHouse | ClickHouse | 657,276,693 | 12,424 | alexey-milovidov | This also works fine:
```
$ curl http://localhost:8123/ --data-binary "SELECT DISTINCT blockSize() FROM numbers(1000000) SETTINGS max_block_size = 10000"
10000
``` | 2020-07-12T21:19:01 | 2020-07-12T21:19:01 | {} | MEMBER |
ClickHouse | ClickHouse | 656,839,487 | 12,294 | alesapin | 
Better | 2020-07-10T19:05:48 | 2020-07-10T19:05:48 | {} | MEMBER |
ClickHouse | ClickHouse | 656,840,808 | 10,230 | alexey-milovidov | You can use the official pre-built binary for FreeBSD:
https://clickhouse.tech/docs/en/getting-started/install/#from-binaries-non-linux
| 2020-07-10T19:07:08 | 2020-07-10T19:07:08 | {
"+1": 2
} | MEMBER |
ClickHouse | ClickHouse | 656,851,687 | 11,568 | alexey-milovidov | Could you please also share the info on how to access Huawei ARM Cloud?
Where it is available, what kind of instance should I order to test this issue?
Also you can send credentials for ssh access to `clickhouse-feedback@yandex-team.com` and we will try to investigate this issue. | 2020-07-10T19:27:13 | 2020-07-10T19:27:13 | {} | MEMBER |
ClickHouse | ClickHouse | 657,128,863 | 11,815 | alexey-milovidov | My colleague @qoega has prepared a dashboard for failed tests. This dashboard is going to become public. | 2020-07-11T20:58:40 | 2020-07-11T20:58:58 | {
"hooray": 1
} | MEMBER |
ClickHouse | ClickHouse | 657,254,741 | 5,462 | alexey-milovidov | > Yes, however storing the value as a String uses less space than as a FixedString(16) if the data is predominantly IPv4.
That's only partially correct.
FixedString(16) occupies 16 bytes in memory.
String with 4 bytes will store 4 bytes + 1 trailing zero byte + 8 byte size = 13 bytes.
That's only 3 bytes differ... | 2020-07-12T17:52:12 | 2020-07-12T17:53:19 | {} | MEMBER |
ClickHouse | ClickHouse | 657,088,451 | 12,342 | genzgd | One more possibly relevant setting: `parallel_view_processing │ 1`
Still working on trying to create a reproducible example | 2020-07-11T16:15:58 | 2020-07-11T16:15:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,172,110 | 12,278 | alexey-milovidov | I see no difference in clang release build for "parallel final" test. | 2020-07-12T04:05:15 | 2020-07-12T04:05:15 | {} | MEMBER |
ClickHouse | ClickHouse | 656,839,378 | 9,461 | alexey-milovidov | The only workaround is to use recent version of libc++.
Either provided in `contrib` submodule (recommended) or installed manually. | 2020-07-10T19:05:42 | 2020-07-10T19:05:42 | {} | MEMBER |
ClickHouse | ClickHouse | 656,886,245 | 10,230 | drook | Wow, thanks a lot, this is like the coolest ever thing to happen to the CH FreeBSD support ! Thanks a lot ! | 2020-07-10T20:55:21 | 2020-07-10T20:55:21 | {} | NONE |
ClickHouse | ClickHouse | 656,937,349 | 12,415 | den-crane | You can implement idempotent inserts and get consistent tables with retries against replicated tables.
But it's tricky. Check this https://clickhouse.tech/docs/en/operations/settings/settings/#settings-deduplicate-blocks-in-dependent-materialized-views | 2020-07-10T23:29:18 | 2020-07-10T23:29:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,266,732 | 11,815 | alexey-milovidov | Sporadic failure in build:
> Error response from daemon: Get https://registry-1.docker.io/v2/yandex/clickhouse-deb-builder/manifests/latest: Get https://auth.docker.io/token?scope=repository%3Ayandex%2Fclickhouse-deb-builder%3Apull&service=registry.docker.io: EOF
2020-07-12 21:59:05,774 Cannot pull image yandex/cli... | 2020-07-12T19:44:56 | 2020-07-12T19:45:13 | {} | MEMBER |
ClickHouse | ClickHouse | 656,835,708 | 8,678 | alexey-milovidov | Ref: #9976 | 2020-07-10T19:02:12 | 2020-07-10T19:02:12 | {} | MEMBER |
ClickHouse | ClickHouse | 656,857,049 | 10,536 | alexey-milovidov | ```
CREATE TABLE mt
(
`x` UInt64 CODEC(NONE, ZSTD(1))
)
ENGINE = MergeTree
ORDER BY tuple()
Received exception from server (version 20.6.1):
Code: 36. DB::Exception: Received from localhost:9000. DB::Exception: It does not make sense to have codec NONE along with other compression codecs: NONE, ZSTD(1).... | 2020-07-10T19:41:23 | 2020-07-10T19:41:23 | {} | MEMBER |
ClickHouse | ClickHouse | 656,859,113 | 11,000 | alexey-milovidov | @4ertus2 This is still relevant. | 2020-07-10T19:46:29 | 2020-07-10T19:46:29 | {} | MEMBER |
ClickHouse | ClickHouse | 656,951,834 | 11,372 | alexey-milovidov | Duplicate of #10470 | 2020-07-11T00:42:09 | 2020-07-11T00:42:09 | {} | MEMBER |
ClickHouse | ClickHouse | 657,102,190 | 12,419 | den-crane | yes such optimization is not implemented
duplicate https://github.com/ClickHouse/ClickHouse/issues/658 | 2020-07-11T17:46:27 | 2020-07-11T17:46:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 656,848,504 | 8,026 | alexey-milovidov | @dmitko Is license the only concern?
We use it with full compliance with the license.
So the issue exists only in the case when all of the following conditions are met:
- you have created and redistribute your own modification of ClickHouse;
- without providing access to it's source code;
- with the SASL library... | 2020-07-10T19:19:19 | 2020-07-10T19:19:19 | {} | MEMBER |
ClickHouse | ClickHouse | 656,852,889 | 2,865 | alexey-milovidov | https://clickhouse.tech/
See "Quick Start", CentOS/RedHat.
CentOS 6 is Ok. | 2020-07-10T19:30:11 | 2020-07-10T19:30:11 | {} | MEMBER |
ClickHouse | ClickHouse | 656,865,600 | 11,124 | alexey-milovidov | This is considered as the expected behaviour.
You have specified that table data is streamed from pipe. You cannot read this data twice.
But we can do better:
- check if it's a regular file or pipe;
- if it's a regular file, seek to begin before read;
- if it's pipe, throw exception if we are trying to read the ... | 2020-07-10T20:03:09 | 2020-07-10T20:03:09 | {} | MEMBER |
ClickHouse | ClickHouse | 656,872,045 | 8,057 | alexey-milovidov | But you read from subquery, not from storage.
The argument `storage` is not "transitive", it applies only to outer SELECT. | 2020-07-10T20:19:41 | 2020-07-10T20:19:41 | {} | MEMBER |
ClickHouse | ClickHouse | 657,257,899 | 12,424 | alexey-milovidov | 
| 2020-07-12T18:23:55 | 2020-07-12T18:23:55 | {} | MEMBER |
ClickHouse | ClickHouse | 657,276,850 | 12,340 | vitlibar | The following syntax is supported (https://github.com/ClickHouse/ClickHouse/pull/11670):
```
DROP ROW POLICY name1 ON [db1.]table_name1 [, name2 ON [db2.]table_name2 ...]
DROP ROW POLICY name1 [, name2 ...] ON [db.]table_name
DROP ROW POLICY name ON [db1.]table_name1 [, [db2.]table_name2 ...]
```
So you can w... | 2020-07-12T21:20:22 | 2020-07-12T21:20:22 | {} | MEMBER |
ClickHouse | ClickHouse | 656,861,379 | 5,264 | alexey-milovidov | Fixed in #10367 since version 20.5. | 2020-07-10T19:52:22 | 2020-07-10T19:52:33 | {} | MEMBER |
ClickHouse | ClickHouse | 656,952,226 | 12,411 | alexey-milovidov | > Functional stateful tests (address) — No output log
```
+ /s3downloader --dataset-names hits visits
INFO:root:Will fetch following datasets: hits, visits
INFO:root:Processing hits
INFO:root:Downloading from https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar to temp path /tmp/dtUqRs
INFO:roo... | 2020-07-11T00:44:38 | 2020-07-11T00:44:38 | {} | MEMBER |
ClickHouse | ClickHouse | 656,970,836 | 12,372 | alexey-milovidov | > Yandex synchronization check (only for Yandex employees)
It's a bug in Arcadia Devtools @exprmntr | 2020-07-11T02:31:36 | 2020-07-11T02:31:36 | {} | MEMBER |
ClickHouse | ClickHouse | 656,971,235 | 12,277 | bobrik | I've added support for PK prefix matches and reorganized code a bit to increase code reuse.
Tests should be happy now, the problem was twofold:
* Missing support for monotonic functions — I don't fully understand how to make it work, made it fall back for now.
* Missing pass-by-reference to `may_be_true_in_range... | 2020-07-11T02:34:34 | 2020-07-11T02:34:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 657,199,242 | 12,409 | mbtolou | thanks
> I meant having `key` be a tuple type column:
>
> ```sql
> create table jtest(
> key Tuple(String, String),
> value Float64
> ) engine=Join(any, left, key);
> insert into jtest values(('a', 'b'), 15)
> select joinGet(jtest, 'value', ('a', 'b'))
> => 15
> ```
>
> Might not work for all u... | 2020-07-12T09:49:39 | 2020-07-12T09:49:39 | {} | NONE |
ClickHouse | ClickHouse | 657,273,730 | 7,895 | alexey-milovidov | The minimum size of the state for `uniqExact` is:
`3 * sizeof(ptr) + 128` = 152 bytes.
128 is the size of hash table to accomodate 8 elements (the hash table of 16 cells 8 bytes each).
We have this for "automatic memory" to save allocations for small states.
For 100 000 000 elements, the amount of memory for st... | 2020-07-12T20:54:34 | 2020-07-12T20:54:53 | {} | MEMBER |
ClickHouse | ClickHouse | 656,852,550 | 2,865 | alexey-milovidov | We have rewritten our RPM packages and there is no such issue anymore. | 2020-07-10T19:29:21 | 2020-07-10T19:29:21 | {} | MEMBER |
ClickHouse | ClickHouse | 656,856,010 | 3,776 | alexey-milovidov | You can pass it as abstract (IColumn) argument and extract the position with `IColumn::getUInt64` virtual function.
(no need to dispatch over column types manually) | 2020-07-10T19:38:30 | 2020-07-10T19:38:30 | {} | MEMBER |
ClickHouse | ClickHouse | 656,860,200 | 11,853 | alexey-milovidov | I have no idea what it is... | 2020-07-10T19:49:16 | 2020-07-10T19:49:16 | {} | MEMBER |
ClickHouse | ClickHouse | 656,883,391 | 8,316 | alexey-milovidov | It may happen if you are upgrading from previously installed unofficial package (that was not asking for password).
So the fact that the password was already specified is not saved and it asks for password on upgrade. | 2020-07-10T20:48:45 | 2020-07-10T20:48:45 | {} | MEMBER |
ClickHouse | ClickHouse | 656,894,094 | 9,248 | alexey-milovidov | It looks like wasted if I test with data size 4000000000 as in your example. Need to investigate... | 2020-07-10T21:14:41 | 2020-07-10T21:14:41 | {} | MEMBER |
ClickHouse | ClickHouse | 657,179,132 | 4,645 | alexey-milovidov | Could you please provide more details? The description is unclear for me.
If you create Log tables on all hosts and query Distributed table, everything should work fine as expected. | 2020-07-12T05:54:19 | 2020-07-12T05:54:19 | {} | MEMBER |
ClickHouse | ClickHouse | 657,257,420 | 12,402 | qoega | It generates several table types. It is trivial to disable Log/TinyLog etc. | 2020-07-12T18:18:53 | 2020-07-12T18:18:53 | {} | MEMBER |
ClickHouse | ClickHouse | 656,834,700 | 7,923 | alexey-milovidov | The old tcmalloc is considered obsolete when there is a new version: https://github.com/ClickHouse/ClickHouse/pull/11590
So, there is no reason to revive the support for using old tcmalloc in ClickHouse. | 2020-07-10T19:01:14 | 2020-07-10T19:01:14 | {} | MEMBER |
ClickHouse | ClickHouse | 656,857,931 | 3,371 | alexey-milovidov | This is how aliases work in ClickHouse, discussion: #9715 | 2020-07-10T19:43:46 | 2020-07-10T19:43:46 | {} | MEMBER |
ClickHouse | ClickHouse | 656,866,061 | 9,102 | alexey-milovidov | #11590 | 2020-07-10T20:04:19 | 2020-07-10T20:04:19 | {} | MEMBER |
ClickHouse | ClickHouse | 656,919,187 | 12,409 | adrian17 | I meant having `key` be a tuple type column:
```sql
create table jtest(
key Tuple(String, String),
value Float64
) engine=Join(any, left, key);
insert into jtest values(('a', 'b'), 15)
select joinGet(jtest, 'value', ('a', 'b'))
=> 15
```
Might not work for all use cases, but for "joinGet with a co... | 2020-07-10T22:20:37 | 2020-07-10T22:20:37 | {} | NONE |
ClickHouse | ClickHouse | 656,846,043 | 8,470 | alexey-milovidov | You can download pre-built binaries for Mac OS X: https://clickhouse.tech/docs/en/getting-started/install/#from-binaries-non-linux | 2020-07-10T19:13:15 | 2020-07-10T19:13:15 | {} | MEMBER |
ClickHouse | ClickHouse | 656,847,981 | 12,372 | s-mx | Hi @qoega! Could you clarify what did I miss in documentation of new setting? | 2020-07-10T19:17:58 | 2020-07-10T19:17:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 656,856,846 | 10,536 | alexey-milovidov | Solved in different way. | 2020-07-10T19:40:47 | 2020-07-10T19:40:47 | {} | MEMBER |
ClickHouse | ClickHouse | 656,866,350 | 9,102 | alexey-milovidov | It's complicated. Currently the new tcmalloc does not provide any performance benefits, only regressions.
| 2020-07-10T20:05:03 | 2020-07-10T20:05:03 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 656,885,366 | 7,966 | alexey-milovidov | There is a tool named `clickhouse-extract-from-config`.
You can use it to preprocess configs and extract the value of some key.
Example:
```
$ clickhouse-extract-from-config --config-file /etc/clickhouse-server/config.xml --key path
/opt/clickhouse/
```
It will also validate config.
You can apply this too... | 2020-07-10T20:53:27 | 2020-07-10T20:53:27 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 656,891,975 | 9,248 | alexey-milovidov | I have tested with data size 100000000 in a loop with 1000 iterations and cannot confirm the leak.
RSS is constant around 2..3 GiB. | 2020-07-10T21:09:28 | 2020-07-10T21:15:18 | {} | MEMBER |
ClickHouse | ClickHouse | 656,952,355 | 11,815 | alexey-milovidov | Connection reset by peer while downloading test data from s3: #12411
Need to add retries. | 2020-07-11T00:45:31 | 2020-07-11T00:45:31 | {} | MEMBER |
ClickHouse | ClickHouse | 657,255,268 | 12,424 | alexey-milovidov | > Performance — 13 faster, 2 slower, 209 unstable
It is a clear win.
But paradoxically, most of the queries that become faster have fixed block_size specified in SETTINGS clause. @akuzm | 2020-07-12T17:57:46 | 2020-07-12T18:09:59 | {} | MEMBER |
ClickHouse | ClickHouse | 657,275,246 | 12,424 | alexey-milovidov | > P.S. I've looked through the code and looks like DEFAULT_BLOCK_SIZE is not used directly (at least for the query above), only via max_block_size
Yes, I have also checked it.
> So after looking into the query_log from the perf tests, it seems that the settings was not applied
But this simple test passes:
```... | 2020-07-12T21:07:33 | 2020-07-12T21:07:33 | {} | MEMBER |
ClickHouse | ClickHouse | 651,953,130 | 12,023 | filimonov | Yes. | 2020-06-30T18:01:33 | 2020-06-30T18:01:33 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 652,851,850 | 11,977 | KochetovNicolai | It will be fixed in #12060 | 2020-07-02T08:00:28 | 2020-07-02T08:00:28 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.