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 | 575,920,869 | 8,650 | alexey-milovidov | Now we can replace std::shared_ptr to boost::intrusive_ptr! | 2020-01-18T17:29:32 | 2020-01-18T17:29:32 | {} | MEMBER |
ClickHouse | ClickHouse | 575,953,227 | 8,717 | oandrew | p.s. MemoryTracker issue happens only on Select. Inserts are ok | 2020-01-19T00:48:37 | 2020-01-19T00:48:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,956,577 | 8,717 | alexey-milovidov | Without compression it is almost the same as TSV:
```
~/ClickHouse$ time clickhouse-client --query "SELECT * FROM test.hits" --format Avro --output_format_avro_codec null > hits.avro.null
real 0m44.714s
user 0m38.064s
sys 0m6.204s
```
But is more compact:
```
~/ClickHouse$ ls -l hits.avro hits.... | 2020-01-19T01:42:34 | 2020-01-19T01:42:34 | {} | MEMBER |
ClickHouse | ClickHouse | 576,193,216 | 5,485 | leoones | @den-crane @vincenzon
I want to subtract the values specified in the array，
I want to find the value of 1-2 and 1-3，
┌─k─v---
│ a │ 1 │
│ a │ 2 │
│ a │ 3 │
| 2020-01-20T09:48:15 | 2020-01-20T09:48:15 | {} | NONE |
ClickHouse | ClickHouse | 576,335,676 | 7,611 | den-crane | @compasses do you speed up single query? or multiple parallel queries? | 2020-01-20T16:00:04 | 2020-01-20T16:00:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,389,032 | 8,751 | 4ertus2 | duplicate #7896 | 2020-01-20T18:34:03 | 2020-01-20T18:34:03 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,953,147 | 8,717 | alexey-milovidov | "unbundled" test fail: as expected (will exclude in CI) | 2020-01-19T00:47:48 | 2020-01-19T00:47:48 | {} | MEMBER |
ClickHouse | ClickHouse | 575,956,821 | 8,717 | oandrew | Yeah I had the same question about calling it `null`. It's from the avro spec.
+1 for `none`. (we can support both `null` and `none` but mention only `none` in the docs/cli help) | 2020-01-19T01:46:56 | 2020-01-19T01:46:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,957,292 | 8,402 | alexey-milovidov | Tests have failed for reason. | 2020-01-19T01:54:47 | 2020-01-19T01:54:47 | {} | MEMBER |
ClickHouse | ClickHouse | 576,005,004 | 8,492 | zhicwu | FYI, I created a repo at https://github.com/zhicwu/clickhouse-datasource-bridge to demonstrate my thoughts. Will add more after holidays. | 2020-01-19T13:32:19 | 2020-01-19T13:32:19 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,774,867 | 8,706 | jmakov | And inserting `'abcd-aa-aa 10:00:11'` makes CK write `1970-01-01 10:00:11`. | 2020-01-17T20:03:28 | 2020-01-17T20:03:28 | {} | NONE |
ClickHouse | ClickHouse | 575,952,356 | 8,717 | alexey-milovidov | > I have tried integrating libserdes (https://github.com/confluentinc/libserdes) locally and it works fine.
Current AvroConfluent implementation is very simple and there is no need to add this library. Is there any other benefits that this library provides?
> I tried adding per-url cache by using LRUCache as a s... | 2020-01-19T00:35:11 | 2020-01-19T00:36:26 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 575,955,207 | 8,717 | alexey-milovidov | But we can also support schema conversion at write (writing data with defined schema).
(We do it for Protobuf) | 2020-01-19T01:19:09 | 2020-01-19T01:20:49 | {} | MEMBER |
ClickHouse | ClickHouse | 575,956,233 | 8,717 | oandrew | Try without compression `--output_format_avro_codec null` | 2020-01-19T01:36:13 | 2020-01-19T01:36:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,043,393 | 8,733 | laplab | Small benchmark for query `SELECT count(*) FROM remote('127.0.0.{1..255}', system.one);`
### This PR
```
Queries executed: 50.
localhost:9000, queries 50, QPS: 2.154, RPS: 549.397, MiB/s: 0.001, result RPS: 2.154, result MiB/s: 0.000.
0.000% 0.406 sec.
10.000% 0.446 sec.
20.000% 0.447 sec.
30.000... | 2020-01-19T20:27:33 | 2020-01-19T20:27:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,073,216 | 8,733 | alexey-milovidov | It is slower on real network requests.
Was:
```
localhost:9000, queries 87, QPS: 4.681, RPS: 1123.360, MiB/s: 0.001, result RPS: 4.681, result MiB/s: 0.000.
0.000% 0.112 sec.
10.000% 0.114 sec.
20.000% 0.114 sec.
30.000% 0.114 sec.
40.000% 0.115 sec.
50.000% ... | 2020-01-20T01:31:35 | 2020-01-20T01:31:35 | {} | MEMBER |
ClickHouse | ClickHouse | 576,255,184 | 5,518 | alexey-milovidov | We have inefficiencies in index analysis.
If we don't count the time spent on index analysis, index query should not be slower than full scan in ClickHouse. | 2020-01-20T12:35:31 | 2020-01-20T12:35:31 | {} | MEMBER |
ClickHouse | ClickHouse | 575,775,089 | 8,704 | alexey-milovidov | Ok but please prepare the description. | 2020-01-17T20:04:02 | 2020-01-17T20:04:02 | {} | MEMBER |
ClickHouse | ClickHouse | 575,836,778 | 4,860 | alexey-milovidov | `DateTime64` with custom resolution has been implemented.
Support for values outside of 1970..2016 is still to do. | 2020-01-17T23:43:58 | 2020-01-17T23:43:58 | {
"heart": 1,
"hooray": 2,
"rocket": 2
} | MEMBER |
ClickHouse | ClickHouse | 575,850,823 | 8,710 | alexey-milovidov | BTW, the function `isFinite` is not optimized to return constant result for integers. | 2020-01-18T01:11:54 | 2020-01-18T01:11:54 | {} | MEMBER |
ClickHouse | ClickHouse | 575,953,766 | 8,717 | alexey-milovidov | > But that should be easy to implement.
Or just postpone the implementation before first use case will appear. | 2020-01-19T00:56:38 | 2020-01-19T00:56:38 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 575,955,362 | 8,717 | alexey-milovidov | > output as-is (same bit representation) - large numbers appear as negative numbers
This. | 2020-01-19T01:21:17 | 2020-01-19T01:21:17 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 575,956,671 | 8,717 | alexey-milovidov | Is `null` the preferred name for "no compression"?
I'm asking because we already have format Null that means - don't write anything; and we already have option to choose generic compression method for text files, and in that option, no compression is `none` not `null`. Maybe it's reasonable to remane `null` to `none` ... | 2020-01-19T01:44:07 | 2020-01-19T01:45:11 | {} | MEMBER |
ClickHouse | ClickHouse | 576,243,138 | 8,112 | alexey-milovidov | ```
SELECT roundBankers(45., -1)
┌─roundBankers(45., -1)─┐
│ 40 │
└───────────────────────┘
``` | 2020-01-20T11:58:21 | 2020-01-20T11:58:21 | {} | MEMBER |
ClickHouse | ClickHouse | 576,209,812 | 8,112 | BayoNet | Is this result correct?
```
SELECT roundBankers(45.33, -1)
┌─roundBankers(45.33, -1)─┐
│ 50 │
└─────────────────────────┘
``` | 2020-01-20T10:30:10 | 2020-01-20T11:58:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,897,374 | 8,624 | nikvas0 | > > Reinventing aerospike (non distributed) ? :)
>
> rocksdb?
I think I should also write rocksdb-based implementation and compare with this dictionary.
Maybe RocksDB with FIFO compaction + lightweight compactions + disabled WAL will be a good solution. Also, there is a table format (https://github.com/faceboo... | 2020-01-18T13:12:45 | 2020-01-18T13:12:45 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,953,606 | 8,717 | oandrew | libserdes has only one additional feature - support for multiple comma separated schema registry URLs. But that should be easy to implement.
https://github.com/ClickHouse/ClickHouse/issues/5601#issuecomment-574566108 | 2020-01-19T00:54:26 | 2020-01-19T00:55:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,954,613 | 8,717 | alexey-milovidov | I mean - type conversion with the same bit representation.
Also UInt8 <-> bool is questionable choice, because UInt8 is often used for small identifiers. | 2020-01-19T01:11:05 | 2020-01-19T01:11:05 | {} | MEMBER |
ClickHouse | ClickHouse | 576,004,497 | 8,726 | alexey-milovidov | Can we do the following:
1. Generate `password_double_sha1_hex` in the same way as `password_sha256_hex` in ClickHouse installation script.
2. Enable `mysql_port` by default.
? | 2020-01-19T13:26:05 | 2020-01-19T13:26:14 | {} | MEMBER |
ClickHouse | ClickHouse | 576,293,872 | 4,470 | smakhtin | Same here - Debian 10 | 2020-01-20T14:17:23 | 2020-01-20T14:17:23 | {} | NONE |
ClickHouse | ClickHouse | 575,554,786 | 2,278 | alexey-milovidov | 19.14
https://github.com/ClickHouse/ClickHouse/blob/master/CHANGELOG.md#clickhouse-release-191433-2019-09-10
> Allow to use multiple threads during parts loading and removal.
| 2020-01-17T09:51:35 | 2020-01-17T09:51:35 | {
"heart": 1
} | MEMBER |
ClickHouse | ClickHouse | 575,832,966 | 8,693 | alexey-milovidov | The table lists children nodes at the path provided.
If it will return empty dataset for non-existing node, how to distinguish it with the case of empty list of children?
Right now the behaviour directly resembles what is happening with the corresponding ZooKeeper request. We request the list of children and ZooKee... | 2020-01-17T23:24:45 | 2020-01-17T23:25:24 | {} | MEMBER |
ClickHouse | ClickHouse | 575,937,030 | 8,571 | alexey-milovidov | Doing HTTP request when parsing format is very strange and impose vulnerabilities.
Do you remember XXE vulnerabilities in XML parsers?
https://owasp.org/www-community/vulnerabilities/XML_External_Entity_(XXE)_Processing
I will check how we can ensure that `RemoteHostFilter` is available and used. | 2020-01-18T20:47:40 | 2020-01-18T20:47:40 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 575,954,403 | 8,717 | alexey-milovidov | Maybe better to support implicit signed <-> unsigned conversions? | 2020-01-19T01:07:25 | 2020-01-19T01:07:25 | {} | MEMBER |
ClickHouse | ClickHouse | 575,955,139 | 8,717 | alexey-milovidov | String must be serialized as Bytes. Because our strings can contain arbitrary binary data. | 2020-01-19T01:18:04 | 2020-01-19T01:18:04 | {} | MEMBER |
ClickHouse | ClickHouse | 575,955,527 | 8,717 | oandrew | I mean yeah truncating is dumb cause it looses data.
I'm fine with same bits (we would just need to make it clear in docs).
Using custom type would be nicely compatible with Clickhouse<->Clickhouse but not with other systems.
| 2020-01-19T01:24:17 | 2020-01-19T01:24:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,957,644 | 8,066 | alexey-milovidov | Unreleased. | 2020-01-19T02:00:16 | 2020-01-19T02:00:16 | {} | MEMBER |
ClickHouse | ClickHouse | 575,958,879 | 8,698 | taiyang-li | The Two failed integration tests seem not caused by my commit. | 2020-01-19T02:21:36 | 2020-01-19T02:21:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,836,081 | 5,301 | alexey-milovidov | My experience:
For development I build debug version with clang on my laptop; I use `ccache` and it helps a lot when switching branches; for performance testing I ssh to one or another development server and build and test there.
In our CI we build on multi-core servers; we use `ccache` and the build time varies ... | 2020-01-17T23:40:20 | 2020-01-17T23:40:20 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 575,837,323 | 4,761 | alexey-milovidov | This will be addressed in the second step of "Polymorphic data parts" feature.
Polymorphic data parts is in development. The first step (compact data parts on disk) is addressed in this PR: https://github.com/ClickHouse/ClickHouse/pull/8290 | 2020-01-17T23:47:13 | 2020-01-17T23:47:13 | {} | MEMBER |
ClickHouse | ClickHouse | 575,890,629 | 8,712 | alexey-milovidov | Now much better:
```
localhost:9000, queries 1, QPS: 1.179, RPS: 117915733.429, MiB/s: 11916.552, result RPS: 11.792, result MiB/s: 0.006.
0.000% 0.848 sec.
10.000% 0.848 sec.
20.000% 0.848 sec.
30.000% 0.848 sec.
40.000% 0.848 sec.
50.000% 0.848 sec.
60.000... | 2020-01-18T11:32:54 | 2020-01-18T11:32:54 | {} | MEMBER |
ClickHouse | ClickHouse | 575,954,509 | 8,717 | oandrew | Avro doesn't support UInt64.
https://avro.apache.org/docs/current/spec.html#schema_primitive
Truncating to Int64 (max avro supports) would mean potentially incorrect data in the output | 2020-01-19T01:09:00 | 2020-01-19T01:09:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,961,720 | 8,568 | huang--yijian | Would you please re-open this issue, as it is not resolved yet it ? | 2020-01-19T03:08:14 | 2020-01-19T03:08:14 | {} | NONE |
ClickHouse | ClickHouse | 576,269,330 | 8,745 | orloffv | The next request is successful.
But after some hours i see this error again | 2020-01-20T13:15:30 | 2020-01-20T13:15:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,317,432 | 8,684 | 4ertus2 | ```
SELECT
item_id,
updated_at
FROM item_distributed ... | 2020-01-20T15:14:25 | 2020-01-20T15:14:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,370,715 | 8,576 | alesapin | Not required in 19.17. | 2020-01-20T17:30:32 | 2020-01-20T17:30:32 | {} | MEMBER |
ClickHouse | ClickHouse | 576,031,152 | 8,730 | alexey-milovidov | ```
CREATE TABLE default.test (foobar.name Array(String), foobar.value Array(String)) ENGINE = Null
```
It's not a valid query. `clickhouse-client` cannot parse it and `clickhouse-server` won't execute.
`clickhouse-format` will display the same error message as the client.
Valid query is shown in the result of... | 2020-01-19T18:19:52 | 2020-01-19T18:19:52 | {} | MEMBER |
ClickHouse | ClickHouse | 575,622,495 | 8,697 | taiyang-li | Something is wrong with this PR, close it now. | 2020-01-17T13:19:23 | 2020-01-17T13:19:23 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,874,750 | 8,701 | alesapin | Seems like we have awful tests for replicated `ALTER` queries :( | 2020-01-18T07:25:28 | 2020-01-18T07:25:28 | {} | MEMBER |
ClickHouse | ClickHouse | 575,919,452 | 8,704 | alexey-milovidov | That's a great feature! | 2020-01-18T17:13:52 | 2020-01-18T17:13:52 | {} | MEMBER |
ClickHouse | ClickHouse | 576,078,106 | 8,569 | dingxiangfei2009 | I should provide more background information. In some cases like network failure, synchronous insertion into a `Distributed` table will fail and silently block the processes. This happens when the thread pool handle is dropped by `DistributedBlockOutputStream`. | 2020-01-20T02:06:37 | 2020-01-20T02:06:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,302,685 | 8,745 | orloffv | problem with compile_expressions and order by | 2020-01-20T14:38:19 | 2020-01-20T14:38:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,371,506 | 8,566 | alesapin | conflict in 19.17 | 2020-01-20T17:32:55 | 2020-01-20T17:32:55 | {} | MEMBER |
ClickHouse | ClickHouse | 576,407,453 | 5,610 | alexey-milovidov | Fixed automatically (by @tavplubix) after enabling `input_format_values_deduce_templates_of_expressions` and `input_format_values_accurate_types_of_literals`. | 2020-01-20T19:40:44 | 2020-01-20T19:40:44 | {} | MEMBER |
ClickHouse | ClickHouse | 575,828,562 | 7,477 | alexey-milovidov | Don't you mind if we steal this task? | 2020-01-17T23:04:52 | 2020-01-17T23:04:52 | {} | MEMBER |
ClickHouse | ClickHouse | 575,956,089 | 8,717 | alexey-milovidov | ```
~/ClickHouse$ time clickhouse-client --query "SELECT * FROM test.hits" --format TSV > hits.tsv
real 0m44.239s
user 0m35.752s
sys 0m5.876s
~/ClickHouse$ time clickhouse-client --query "SELECT * FROM test.hits" --format Avro > hits.avro
real 1m4.151s
user 0m59.476s
sys 0m4.112s
```
... | 2020-01-19T01:33:23 | 2020-01-19T01:33:23 | {} | MEMBER |
ClickHouse | ClickHouse | 576,242,922 | 8,112 | alexey-milovidov | @BayoNet Yes, it is. Because the value is greater than the half (45). | 2020-01-20T11:57:43 | 2020-01-20T11:57:43 | {} | MEMBER |
ClickHouse | ClickHouse | 576,246,043 | 8,742 | alexey-milovidov | We have newer versions, both minor (19.17) and patch (19.16.9). | 2020-01-20T12:06:57 | 2020-01-20T12:06:57 | {} | MEMBER |
ClickHouse | ClickHouse | 576,358,210 | 8,705 | alesapin | Don't need to cherry-pick into 19.17. | 2020-01-20T16:56:17 | 2020-01-20T16:56:17 | {} | MEMBER |
ClickHouse | ClickHouse | 576,371,679 | 8,549 | alesapin | conflict in 19.17 | 2020-01-20T17:33:25 | 2020-01-20T17:33:25 | {} | MEMBER |
ClickHouse | ClickHouse | 576,406,480 | 8,750 | azat | >but now I don't have any ideas why
Got it, I want to have separate default disk path (since default selector uses `<path>` and I want to use `<tmp_path>`) | 2020-01-20T19:37:20 | 2020-01-20T19:37:20 | {} | MEMBER |
ClickHouse | ClickHouse | 576,411,632 | 2,341 | alexey-milovidov | @KochetovNicolai Надо проверить, скорее всего давно исправлено. | 2020-01-20T19:56:21 | 2020-01-20T19:56:21 | {} | MEMBER |
ClickHouse | ClickHouse | 575,954,793 | 8,721 | zainal-abidin-assegaf | https://github.com/Altinity/clickhouse-operator/issues/236 | 2020-01-19T01:14:07 | 2020-01-19T01:14:07 | {} | NONE |
ClickHouse | ClickHouse | 576,006,206 | 8,730 | chhetripradeep | Not sure why a Nested type is shown as Array | 2020-01-19T13:46:40 | 2020-01-19T13:46:40 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,245,724 | 8,742 | alexey-milovidov | It doesn't reproduce on master:
```
milovidov-Pro-P30 :) CREATE TABLE testtable (time DateTime, int_value Int64) ENGINE = MergeTree ORDER BY (time);
CREATE TABLE testtable
(
`time` DateTime,
`int_value` Int64
)
ENGINE = MergeTree
ORDER BY time
Ok.
0 rows in set. Elapsed: 0.012 sec.
milovid... | 2020-01-20T12:05:59 | 2020-01-20T12:05:59 | {} | MEMBER |
ClickHouse | ClickHouse | 576,305,314 | 8,742 | darktemplarbasealt | I've checked version **19.17.6.36** and it no longer crashes for me too. Thank you. | 2020-01-20T14:44:42 | 2020-01-20T14:44:42 | {} | NONE |
ClickHouse | ClickHouse | 576,404,423 | 8,750 | azat | >BTW, do you consider the option to use existing storage_configuration and just to refer to some policy to store temporary data?
At first I went this way (and just has `tmp_policy`), but now I don't have any ideas why, using existing `storage_configuration` looks indeed more convenient. Should be rebased... | 2020-01-20T19:30:33 | 2020-01-20T19:30:33 | {} | MEMBER |
ClickHouse | ClickHouse | 576,024,858 | 8,717 | alexey-milovidov | Current status:
- need to add type conversions in reader, then we can merge.
Next steps (after merge):
- support for `input_format_null_as_default` setting;
- support for schema-aware output;
- enable AvroConfluent with appropriate changes. | 2020-01-19T17:04:35 | 2020-01-19T17:04:35 | {} | MEMBER |
ClickHouse | ClickHouse | 576,092,082 | 8,730 | chhetripradeep | Thank you for describing in detail. | 2020-01-20T03:30:47 | 2020-01-20T03:30:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,399,784 | 8,750 | alexey-milovidov | Ok.
BTW, do you consider the option to use existing `storage_configuration` and just to refer to some policy to store temporary data? | 2020-01-20T19:13:43 | 2020-01-20T19:13:43 | {} | MEMBER |
ClickHouse | ClickHouse | 576,403,605 | 8,736 | alexey-milovidov | @dingxiangfei2009 Could you please check test results? | 2020-01-20T19:27:53 | 2020-01-20T19:27:53 | {} | MEMBER |
ClickHouse | ClickHouse | 575,831,276 | 5,618 | alexey-milovidov | @kainever7 You should set `max_bytes_before_external_group_by` to slightly less value that the total amount of memory. For example, if you have 16 GB of memory, it's reasonable to set `max_memory_usage` to 15000000000 (because some memory is needed for OS and for other programs in the system) and `max_bytes_before_exte... | 2020-01-17T23:16:54 | 2020-01-17T23:17:49 | {} | MEMBER |
ClickHouse | ClickHouse | 575,839,074 | 2,345 | alexey-milovidov | This issue is in queue. The reason is slow index analysis. | 2020-01-17T23:56:37 | 2020-01-17T23:56:37 | {} | MEMBER |
ClickHouse | ClickHouse | 575,943,195 | 8,717 | oandrew | @alexey-milovidov
Disabling AvroConfluent makes sense - it needs more work.
I have tried integrating libserdes (https://github.com/confluentinc/libserdes) locally and it works fine.
My only concern about libserdes is that it brings in additional dependencies - `curl` and `libjansson`. Also I had make some chan... | 2020-01-18T22:14:47 | 2020-01-18T23:40:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,999,249 | 8,719 | alexey-milovidov | https://clickhouse-test-reports.s3.yandex.net/8719/28ad3b46da493b7abdcf2f4e9884971577820580/performance_test/comparison_max_rps_gcc_9.html
5 times better. | 2020-01-19T12:22:46 | 2020-01-19T12:22:46 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 576,073,944 | 8,733 | alexey-milovidov | With non-zero queries it is also 100ms slower.
Need to parallelize connection and sending queries. | 2020-01-20T01:37:01 | 2020-01-20T01:37:01 | {} | MEMBER |
ClickHouse | ClickHouse | 576,179,091 | 8,159 | Aaronzk | 两位大神，如果有结果了，请分享一下完整的执行步骤，万分感谢。 | 2020-01-20T09:13:52 | 2020-01-20T09:13:52 | {} | NONE |
ClickHouse | ClickHouse | 576,192,955 | 8,693 | bgranvea | ok that makes sense | 2020-01-20T09:47:37 | 2020-01-20T09:47:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,195,563 | 8,739 | godfreyd | @4ertus2, посмотри, плиз. Ты в этом хорошо разбираешься. | 2020-01-20T09:54:15 | 2020-01-20T09:54:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,280,995 | 8,684 | m1khal3v | The problem is not in the version. Transferred all the data to version 19.15.2 got absolutely identical behavior. Before that, there was a different data set, but the structure was identical. However, it turned out that the problem is reproduced only when specific two tables are joined, that is, if you replace item_dis... | 2020-01-20T13:45:15 | 2020-01-20T14:09:41 | {} | NONE |
ClickHouse | ClickHouse | 576,356,677 | 8,749 | ikopylov | @alexey-milovidov PTAL. I've added special handling for Int16 and Int8 types to avoid slow `bit_cast` call for all integer types | 2020-01-20T16:52:07 | 2020-01-20T16:52:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,412,049 | 2,592 | alexey-milovidov | @Potya | 2020-01-20T19:58:06 | 2020-01-20T19:58:06 | {} | MEMBER |
ClickHouse | ClickHouse | 575,742,124 | 8,703 | alexey-milovidov | Can be solved when we will have information of data properties on "ports" of "processors". | 2020-01-17T18:30:03 | 2020-01-17T18:30:03 | {} | MEMBER |
ClickHouse | ClickHouse | 575,937,113 | 8,571 | alexey-milovidov | And we don't setup any timeouts for HTTP request...
You know - any request without timeout is anti-pattern. | 2020-01-18T20:48:44 | 2020-01-18T20:48:54 | {
"+1": 2
} | MEMBER |
ClickHouse | ClickHouse | 575,955,340 | 8,717 | oandrew | For UInt64 (and UInt32) I see few options:
- truncate to Int64 - large numbers truncated as 2^63-1
- output as-is (same bit representation) - large numbers appear as negative numbers
- have a special flag that enables one of the above
- output as custom type fixed 8-bytes. << I think I prefer this one the most | 2020-01-19T01:21:00 | 2020-01-19T01:21:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,972,389 | 8,728 | Akazz | N.B. This PR is recommended to be backported alongside with #8718! | 2020-01-19T06:35:48 | 2020-01-19T06:35:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,090,805 | 8,583 | kuncao | from the expert Alexander Zaitsev ,who works in altinity.com,our issue is :
1. For sum() you do not need to use AggregateState (sumState) functions at all. SummingMergeTree would sum automatically. If you are using SummingMergeTree please change sumState() to sum(),
2. Alternativelly, you could use AggregatingMergeT... | 2020-01-20T03:23:27 | 2020-01-20T03:23:51 | {} | NONE |
ClickHouse | ClickHouse | 575,628,905 | 8,695 | alex-krash | I've finished with initial implementation, waiting for review/comments, etc. After done, will implement functional test and documentation. | 2020-01-17T13:37:48 | 2020-01-17T13:37:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,919,713 | 8,691 | alexey-milovidov | Ok but `round` is also using bankers rounding for floating point numbers because of default rounding mode of CPU. The difference is for integers that we can round to specified power of ten. Please do experiments and clarify... | 2020-01-18T17:17:04 | 2020-01-18T17:17:50 | {} | MEMBER |
ClickHouse | ClickHouse | 576,253,593 | 7,611 | compasses | > Personally I don't recommend this schema because second replicas kill performance (they use cache / I/O/ CPU).
Your mean each shard has one replicas, and in single one node have two shards, may have concurrent query in the same time, right?
But, I'm very confused about the performance concern here, we use shard... | 2020-01-20T12:30:18 | 2020-01-20T12:30:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,359,948 | 8,672 | alesapin | conflict in 19.17 | 2020-01-20T17:00:58 | 2020-01-20T17:00:58 | {} | MEMBER |
ClickHouse | ClickHouse | 576,401,727 | 5,671 | alexey-milovidov | CPU load is high for two reasons:
- there are only 8 cores on a server;
- data is highly compressable and large amount of CPU load is used for compression and decompression;
To mitigate the first issue you can lower the `background_pool_size` setting for profile `default` in `users.xml`. Recommended value is 8.
... | 2020-01-20T19:20:57 | 2020-01-20T19:21:32 | {} | MEMBER |
ClickHouse | ClickHouse | 576,407,556 | 5,610 | alexey-milovidov | I will add your use case as a test. | 2020-01-20T19:41:11 | 2020-01-20T19:41:11 | {} | MEMBER |
ClickHouse | ClickHouse | 575,678,682 | 8,263 | alexey-milovidov | Does it reproduce in 20.2 testing?
https://repo.yandex.ru/clickhouse/deb/testing/main/ | 2020-01-17T15:44:41 | 2020-01-17T15:44:41 | {} | MEMBER |
ClickHouse | ClickHouse | 575,953,164 | 8,721 | den-crane | You should ask this at https://github.com/Altinity/clickhouse-operator/issues | 2020-01-19T00:48:04 | 2020-01-19T00:48:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 575,953,415 | 6,256 | alexey-milovidov | We need to apply any sort of magic to make the function as fast as before. | 2020-01-19T00:51:35 | 2020-01-19T00:51:35 | {} | MEMBER |
ClickHouse | ClickHouse | 575,954,158 | 8,717 | alexey-milovidov | ```
~/ClickHouse$ time clickhouse-client --query "SELECT * FROM test.hits" --format Avro > hits.avro
Code: 44. DB::Exception: Type UInt64 is not supported for Avro output: column WatchID
```
Why so bad? | 2020-01-19T01:03:33 | 2020-01-19T01:03:33 | {} | MEMBER |
ClickHouse | ClickHouse | 575,956,723 | 8,717 | oandrew | p.s. Avro with snappy also calculates crc (avro cpp uses boost crc for that) and it is not optimized (in case someone wants to speed it up)
also `--output_format_avro_sync_interval` sets the size of compression block. by default it's 16Kb | 2020-01-19T01:45:07 | 2020-01-19T01:53:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 576,246,359 | 8,742 | darktemplarbasealt | Thanks, I'll check it on newer versions. | 2020-01-20T12:07:57 | 2020-01-20T12:07:57 | {} | NONE |
ClickHouse | ClickHouse | 595,212,255 | 9,525 | alexey-milovidov | `for i in {1..100}; do ./clickhouse-test --zookeeper zookeeper_test_alter_compression_codecs; done`
Ok in debug build. | 2020-03-05T12:50:04 | 2020-03-05T12:50:04 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.