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 | 622,541,542 | 10,585 | alexey-milovidov | At least it works. | 2020-05-01T20:00:21 | 2020-05-01T20:00:21 | {} | MEMBER |
ClickHouse | ClickHouse | 622,550,683 | 8,842 | tavplubix | In #7572 query parameters can be sent through URL parameters or request headers for any HTTP method, but looks like it does not allow sending query parameters through POST body. | 2020-05-01T20:21:14 | 2020-05-01T20:21:14 | {} | MEMBER |
ClickHouse | ClickHouse | 622,581,037 | 10,604 | alexey-milovidov | Still does not work. | 2020-05-01T21:45:41 | 2020-05-01T21:45:41 | {} | MEMBER |
ClickHouse | ClickHouse | 622,980,350 | 10,625 | alexey-milovidov | Now it looks like that gcc and clang are on-par.
Let's see what will change after update to clang-10.
And we should make more careful test with `performance-test`. | 2020-05-02T16:35:22 | 2020-05-02T16:35:22 | {} | MEMBER |
ClickHouse | ClickHouse | 622,982,122 | 10,624 | den-crane | >is there any way in CH while import i can say use backslash as escape character. ?
No. CH supports only CSV with "double quote is escaped by another double quote" | 2020-05-02T16:48:48 | 2020-05-02T16:49:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 623,002,977 | 6,638 | azat | FWIW looks like it works in upstream fc28416b81 (but I guess it is better to add a test from this issue) | 2020-05-02T19:34:02 | 2020-05-02T19:34:02 | {} | MEMBER |
ClickHouse | ClickHouse | 621,694,740 | 10,559 | ildus | That's just awesome. In my benchmarks tuple version works even faster now, and I see performance improvement about 60%. | 2020-04-30T08:34:22 | 2020-04-30T08:34:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,157,278 | 10,473 | nikitamikhaylov | Maybe you can use `clickhouse-local`? | 2020-04-30T22:44:49 | 2020-04-30T22:44:49 | {} | MEMBER |
ClickHouse | ClickHouse | 622,419,463 | 10,604 | vzakaznikov | > @vzakaznikov The same test has failed again.
It failed for a different reason. Looking into it. | 2020-05-01T14:52:17 | 2020-05-01T14:52:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,491,223 | 3,990 | diafour | @gyrter FYI
There was already an attempt to add Clickhouse to helm/charts: https://github.com/helm/charts/pull/6371 with more options in values.yaml and without loghouse specific settings.
| 2020-05-01T17:50:34 | 2020-05-01T17:50:34 | {} | NONE |
ClickHouse | ClickHouse | 622,556,017 | 10,585 | alexey-milovidov | > Some micro-optimizations are required to close the performance gap.
There are many cases when we put functions to headers for inlining.
LTO can eliminate the need for it and find more cases where inlining can improve code. | 2020-05-01T20:33:05 | 2020-05-01T20:33:31 | {} | MEMBER |
ClickHouse | ClickHouse | 622,971,581 | 10,624 | alexey-milovidov | CSV has different escaping rules: double quote is escaped by another double quote, not by backslash.
Example:
```
"Fri, 01 May 2020 06:47:05 UTC","<html id=""html-div"">"
```
| 2020-05-02T15:33:34 | 2020-05-02T15:33:34 | {} | MEMBER |
ClickHouse | ClickHouse | 623,002,351 | 9,354 | azat | FWIW Looks like it works in upstream (v20.4.1.3177-prestable-92-gfc28416b81, fc28416b81) | 2020-05-02T19:28:44 | 2020-05-02T19:28:58 | {} | MEMBER |
ClickHouse | ClickHouse | 623,004,405 | 7,546 | alexey-milovidov | We don't support AppleClang, only clang and gcc. | 2020-05-02T19:46:01 | 2020-05-02T19:46:01 | {} | MEMBER |
ClickHouse | ClickHouse | 623,039,041 | 7,704 | alexey-milovidov | Writing of Array into Parquet is not implemented at this moment. | 2020-05-03T01:29:25 | 2020-05-03T01:29:25 | {} | MEMBER |
ClickHouse | ClickHouse | 621,706,614 | 10,000 | akuzm | Oops, looks like github couldn't detect the merge. I'll close this one manually, hope it doesn't mess up the changelog. | 2020-04-30T08:58:02 | 2020-04-30T08:58:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,790,637 | 10,585 | alexey-milovidov | It looks like we don't have llvm-ar in our Docker images... | 2020-04-30T12:04:46 | 2020-04-30T12:04:46 | {} | MEMBER |
ClickHouse | ClickHouse | 622,511,336 | 10,609 | alexey-milovidov | TSan tests failed for unknown reason. | 2020-05-01T18:42:11 | 2020-05-01T18:42:11 | {} | MEMBER |
ClickHouse | ClickHouse | 622,980,405 | 10,625 | alexey-milovidov | https://raw.githubusercontent.com/ClickHouse/ClickHouse/5a5d64eb90cb877009b576f7d66a56656d7154f2/website/benchmark_gcc_clang.html | 2020-05-02T16:35:46 | 2020-05-02T16:35:46 | {} | MEMBER |
ClickHouse | ClickHouse | 622,482,925 | 10,612 | den-crane | https://clickhouse.tech/docs/en/sql-reference/statements/create/#create-dictionary-query
ClickHouse Release V19.17.4.11, 2019-11-22
New Feature
Add the ability to create dictionaries with DDL queries
And you can create a dictionary using the old way -- xml file
https://clickhouse.tech/docs/en/sql-referenc... | 2020-05-01T17:30:12 | 2020-05-01T17:30:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 623,002,623 | 10,588 | alexey-milovidov | Tests are broken. | 2020-05-02T19:31:04 | 2020-05-02T19:31:04 | {} | MEMBER |
ClickHouse | ClickHouse | 623,004,130 | 4,903 | alexey-milovidov | Current status: ORC output format is still to do. | 2020-05-02T19:43:36 | 2020-05-02T19:43:36 | {} | MEMBER |
ClickHouse | ClickHouse | 623,006,030 | 10,461 | alexey-milovidov | It happens because ParquetBlockInputFormat copies all file content in memory if file is not seekable.
The same happens for ORC reader (but unconditionally). | 2020-05-02T20:00:37 | 2020-05-02T20:00:58 | {} | MEMBER |
ClickHouse | ClickHouse | 621,791,676 | 10,585 | alexey-milovidov | I'll add it. | 2020-04-30T12:06:58 | 2020-04-30T12:06:58 | {} | MEMBER |
ClickHouse | ClickHouse | 621,903,465 | 10,591 | alex-zaitsev | Agree, but it is probably sufficient to have two columns, not 3:
move_ttl_info.name Array(String),
move_ttl_info.ttl_minmax Array((DateTime, DateTime)),
| 2020-04-30T14:51:23 | 2020-04-30T14:51:23 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,467,394 | 10,050 | Phantomape | Any update on this pr？ | 2020-05-01T16:50:41 | 2020-05-01T16:50:41 | {} | NONE |
ClickHouse | ClickHouse | 622,503,789 | 10,277 | alexey-milovidov | `./0_stateless/01012_serialize_array_memory_usage.sql`
`./1_stateful/00154_avro.sql` | 2020-05-01T18:23:00 | 2020-05-01T18:23:00 | {} | MEMBER |
ClickHouse | ClickHouse | 622,685,193 | 10,612 | kitex | >
>
> https://clickhouse.tech/docs/en/sql-reference/statements/create/#create-dictionary-query
>
> ClickHouse Release V19.17.4.11, 2019-11-22
> New Feature
> Add the ability to create dictionaries with DDL queries
>
> And you can create a dictionary using the old way -- xml file
> https://clickhouse.tech/d... | 2020-05-02T06:02:05 | 2020-05-02T06:02:05 | {} | NONE |
ClickHouse | ClickHouse | 623,006,459 | 10,461 | alexey-milovidov | Maybe we can implement `arrow::io::RandomAccessFile` from our ReadBuffer if Arrow library can only seek forward when reading Parquet or ORC.
@oandrew @proller @wesm | 2020-05-02T20:04:43 | 2020-05-02T20:04:43 | {} | MEMBER |
ClickHouse | ClickHouse | 623,020,451 | 10,461 | wesm | You might be interested in our C++ filesystem wrapper for the SDK which implements seek, range reads, and other APIs required for Parquet files
https://github.com/apache/arrow/blob/master/cpp/src/arrow/filesystem/s3fs.cc | 2020-05-02T22:00:25 | 2020-05-02T22:00:25 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 621,730,851 | 4,959 | xmabul | > ClickHouse server version 19.14.6:
> when i use mysql cilent, get this error:
> 2019.11.10 17:15:35.159193 [ 183 ] {} MySQLHandler: DB::Exception: Cannot read all data. Bytes read: 0. Bytes expected: 3.
@dbaops-com hello,I met the same problem.Do u have solved it? | 2020-04-30T09:49:30 | 2020-04-30T09:49:30 | {} | NONE |
ClickHouse | ClickHouse | 621,826,658 | 10,594 | filimonov | And what is in: `/var/log/clickhouse-server/clickhouse-server.err.log`
Do you see any lines with lot of '#####' ?
```
grep '#####' -C 100 /var/log/clickhouse-server/clickhouse-server.err.log
```
| 2020-04-30T13:13:00 | 2020-04-30T13:13:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,877,460 | 9,334 | raimonte | @filimonov
I have a similar problem
Docker settings:
```
services:
clickhouse-server:
image: yandex/clickhouse-server:latest
restart: always
ports:
- 8123:8123
- 9000:9000
ulimits:
nproc: 65535
nofile:
soft: 262144
hard: 262144
kafka:
im... | 2020-04-30T14:10:35 | 2020-04-30T15:54:49 | {} | NONE |
ClickHouse | ClickHouse | 622,381,885 | 10,604 | alexey-milovidov | @vzakaznikov The same test has failed again. | 2020-05-01T13:11:13 | 2020-05-01T13:11:13 | {} | MEMBER |
ClickHouse | ClickHouse | 622,394,514 | 5,747 | amitsudharshan | Does this feature mean using Kerberos for authentication for all ClickHouse ACLs, or is this somehow specific to Zk/hdfs. I’m very interested in Kerberos for all.
Thanks! | 2020-05-01T13:45:00 | 2020-05-01T13:45:00 | {} | NONE |
ClickHouse | ClickHouse | 623,002,744 | 10,580 | alexey-milovidov | Minor:
```
2020-04-29 15:14:08 In file included from ../src/Processors/Formats/Impl/ArrowBlockInputFormat.cpp:1:
2020-04-29 15:14:08 ../src/Processors/Formats/Impl/ArrowBlockInputFormat.h:35:7: error: no newline at end of file [-Werror,-Wnewline-eof]
2020-04-29 15:14:08 #endif
2020-04-29 15:14:08 ^
2020-04-... | 2020-05-02T19:32:01 | 2020-05-02T19:32:01 | {} | MEMBER |
ClickHouse | ClickHouse | 623,015,787 | 10,604 | vzakaznikov | @alexey-milovidov, I think I've beaten these tests into submission. Any way to re-run them a couple of more times?
| 2020-05-02T21:29:41 | 2020-05-02T21:29:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,872,995 | 10,579 | 4ertus2 | Waiting for #10587 | 2020-04-30T14:03:11 | 2020-04-30T14:03:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,086,721 | 10,163 | Avogar | https://github.com/ClickHouse/ClickHouse/pull/10607 | 2020-04-30T20:17:54 | 2020-04-30T20:17:54 | {} | MEMBER |
ClickHouse | ClickHouse | 622,480,687 | 10,612 | kitex | > CREATE DICTIONARY item_dict ( eci_lu Int64, edu_lu Int64, pri_lu Int64, location String )
> PRIMARY KEY eci_lu,edu_lu,pri_lu SOURCE(CLICKHOUSE(HOST 'localhost' PORT 9000
> TABLE b DB 'default' USER 'default'))
> LIFETIME(MIN 0 MAX 0) LAYOUT(COMPLEX_KEY_HASHED());
I could not run the query, is it dictionary ... | 2020-05-01T17:24:30 | 2020-05-01T17:24:30 | {} | NONE |
ClickHouse | ClickHouse | 622,560,872 | 10,585 | amosbird | > There are many cases when we put functions to headers for inlining.
Yeah, but I think we've already covered most inlinable cases that can produce performance benefits.
One thing I noticed can have noticeable effects on performance when doing micro-optimizations is the register allocation algorithm. And another ... | 2020-05-01T20:44:57 | 2020-05-01T20:44:57 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 622,951,600 | 10,621 | azat | >I'm looking into make Aggregator respect column names (if that makes sense?)
e8104d14aa | 2020-05-02T13:12:00 | 2020-05-02T13:12:00 | {} | MEMBER |
ClickHouse | ClickHouse | 622,960,453 | 10,625 | alexey-milovidov | Yes, it gives one percent more :)
Let's also compare all combined together... | 2020-05-02T14:17:20 | 2020-05-02T14:17:20 | {} | MEMBER |
ClickHouse | ClickHouse | 622,981,746 | 10,624 | shivakumarss | I was just going through the CSV specification, there are ways to escape it by either double quotes or by backslash.
https://en.wikipedia.org/wiki/Comma-separated_values
`CSV is a delimited text file that uses a comma to separate values (many implementations of CSV import/export tools allow other separators ... | 2020-05-02T16:45:52 | 2020-05-02T16:45:52 | {} | NONE |
ClickHouse | ClickHouse | 621,885,736 | 10,590 | den-crane | Try `set max_threads=1`
Probably it just not enough memory for `select *` in many threads with max_blocks_size=65536
and the limit indirectly reduces the number of blocks then number of threads.
| 2020-04-30T14:23:33 | 2020-04-30T14:24:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,939,462 | 10,593 | filimonov | Also: it may relate https://github.com/ClickHouse/ClickHouse/issues/9587 | 2020-04-30T15:50:02 | 2020-04-30T15:50:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,429,391 | 2,389 | alexeychernykh | Not reproduced on version 19.17.9.60. Maybe it is fixed. | 2020-05-01T15:17:06 | 2020-05-01T15:17:37 | {} | NONE |
ClickHouse | ClickHouse | 622,436,513 | 10,585 | alexey-milovidov | Looks like it was built successfully:
https://clickhouse-builds.s3.yandex.net/10585/f11148a77c1f387158fe43f3203d47402aee557f/report.html
this is
clang-9 | relwithdebuginfo | none | bundled | unsplitted | 2020-05-01T15:34:23 | 2020-05-01T15:34:23 | {} | MEMBER |
ClickHouse | ClickHouse | 622,440,867 | 10,585 | alexey-milovidov | Looks like the final linking stage took 10 minutes:
```
2020-05-01 14:33:29 [7273/7283] Linking CXX executable programs/clickhouse-odbc-bridge
2020-05-01 14:33:29 ld.lld: warning: <unknown>:0:0: loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled ... | 2020-05-01T15:44:46 | 2020-05-01T15:44:46 | {} | MEMBER |
ClickHouse | ClickHouse | 622,971,809 | 10,624 | alexey-milovidov | > this csv is created using apache spark
It means that either Apache Spark misses CSV specification or that you did not specified it correctly. | 2020-05-02T15:35:27 | 2020-05-02T15:35:27 | {} | MEMBER |
ClickHouse | ClickHouse | 621,726,705 | 10,559 | ildus | Thank you | 2020-04-30T09:40:22 | 2020-04-30T09:40:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,419,744 | 10,560 | Enmk | TODO: after merging, update `WHERE` example in the docs:
https://clickhouse.tech/docs/en/sql-reference/data-types/datetime64/#examples | 2020-05-01T14:52:57 | 2020-05-01T14:52:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,502,311 | 10,277 | alexey-milovidov | Queries from functional tests with more than 1 GiB memory usage:
```
2020.05.01 19:08:58.771440 [ 952478 ] {ac183c16-bd87-4f11-9087-862c66674634} <Debug> executeQuery: (from [::1]:49348) SELECT ignore(x) FROM (SELECT groupArray(number) AS x FROM numbers(33554433)) GROUP BY x FORMAT Null
2020.05.01 19:10:36.847080 [ ... | 2020-05-01T18:19:11 | 2020-05-01T18:19:11 | {} | MEMBER |
ClickHouse | ClickHouse | 622,935,851 | 10,585 | amosbird | > Also remember that we don't omit-frame-pointer in clang builds.
Let's try if it can close the gap now :) | 2020-05-02T11:03:05 | 2020-05-02T11:03:05 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 623,035,396 | 10,604 | vzakaznikov | For some reason these debug runs did not start. I think the CI run needs to be restarted. | 2020-05-03T00:47:17 | 2020-05-03T00:47:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,790,033 | 10,595 | alexey-milovidov | 
| 2020-04-30T12:03:26 | 2020-04-30T12:03:26 | {} | MEMBER |
ClickHouse | ClickHouse | 621,791,112 | 10,585 | alexey-milovidov | Or we don't have appropriate symlink to llvm-ar-10. | 2020-04-30T12:05:47 | 2020-04-30T12:05:47 | {} | MEMBER |
ClickHouse | ClickHouse | 621,887,994 | 10,591 | filimonov | Проблема 1 - как показать move_ttl. Варианты - массив туплов (не self-desctibing) или Nested (не всегда удобный, self-describing, используется в разных местах - например system.query_log.ProfileEvents)
Проблема 2 - как показать ttl для части. Мне кажется просто 2 колонки (ttl_min / ttl_max) в system.parts
Пробле... | 2020-04-30T14:26:57 | 2020-04-30T14:29:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,550,350 | 10,585 | alexey-milovidov | On basic set of queries from https://clickhouse.tech/benchmark/hardware/ there is no visible difference. | 2020-05-01T20:20:27 | 2020-05-01T20:20:27 | {} | MEMBER |
ClickHouse | ClickHouse | 622,940,280 | 10,617 | alexey-milovidov | It did not help. | 2020-05-02T11:39:52 | 2020-05-02T11:39:52 | {} | MEMBER |
ClickHouse | ClickHouse | 623,002,462 | 10,591 | alexey-milovidov | For table TTLs we already have:
TTL expr MOVE
TTL expr DELETE
and there are upcoming
TTL expr DELETE WHERE cond
TTL expr GROUP BY columns... SET col = agg(col), ...
Now suppose we will allow multiple table TTLs.
And we should not limit this new info for TTL MOVE, because it is rarely used feature that ... | 2020-05-02T19:29:39 | 2020-05-02T19:29:39 | {} | MEMBER |
ClickHouse | ClickHouse | 623,022,217 | 10,604 | alexey-milovidov | Ok. For unknown reason I don't see "debug" test runs. Will wait a bit more or re-run. | 2020-05-02T22:12:13 | 2020-05-02T22:12:13 | {} | MEMBER |
ClickHouse | ClickHouse | 622,422,959 | 10,604 | vzakaznikov | @alexey-milovidov, pushed a fix. Let's see how it behaves this time. | 2020-05-01T15:00:57 | 2020-05-01T15:00:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,434,965 | 10,604 | alexey-milovidov | :+1: | 2020-05-01T15:30:37 | 2020-05-01T15:30:37 | {} | MEMBER |
ClickHouse | ClickHouse | 622,509,347 | 8,683 | alexey-milovidov | Implementation is fairly easy. Provide a method in StorageDistributed to get queue size. Add system storage similar to system.replication_queue. It will go through all databases and tables, do dynamic_cast to StorageDistributed and collect the result. | 2020-05-01T18:36:43 | 2020-05-01T18:36:43 | {} | MEMBER |
ClickHouse | ClickHouse | 622,555,350 | 10,585 | amosbird | I guess vectorized processing won't benefit much from cross TU optimizations. Some micro-optimizations are required to close the performance gap. | 2020-05-01T20:31:25 | 2020-05-01T20:31:25 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 622,571,480 | 10,585 | alexey-milovidov | [benchmark_gcc_clang.html.txt](https://github.com/ClickHouse/ClickHouse/files/4566306/benchmark_gcc_clang.html.txt)
ThinLTO can cut half of the gap between gcc and clang.
We need to... | 2020-05-01T21:15:19 | 2020-05-01T21:15:19 | {} | MEMBER |
ClickHouse | ClickHouse | 622,573,130 | 10,585 | alexey-milovidov | Also remember that we don't omit-frame-pointer in clang builds. | 2020-05-01T21:20:14 | 2020-05-01T21:20:14 | {} | MEMBER |
ClickHouse | ClickHouse | 622,955,770 | 10,625 | alexey-milovidov | Tests have passed, let's look at the performance... | 2020-05-02T13:43:00 | 2020-05-02T13:43:00 | {} | MEMBER |
ClickHouse | ClickHouse | 623,016,480 | 10,437 | rsurgiewicz | I'm having same issue,
I have
1. table A engine ReplaceMergeTree()
2. MV B on table A
3. Buffer table on table A
Data is not written to the destination A table from the buffer.
Even DROPing the Buffer that should trigger flush the data fails
```
2020.05.02 21:31:28.038219 [ 526 ] {06a0c14c-6c81-4204-9ab7-9... | 2020-05-02T21:36:57 | 2020-05-02T21:37:41 | {} | NONE |
ClickHouse | ClickHouse | 621,705,007 | 10,559 | akuzm | Good to know, I'll merge it then. | 2020-04-30T08:55:02 | 2020-04-30T08:55:02 | {
"hooray": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,275,180 | 7,287 | doom369 | Same issue here:
```
2020.05.01 02:17:04.277285 [ 30977 ] {} <Error> void DB::BackgroundProcessingPool::threadFunction(): Code: 49, e.displayText() = DB::Exception: Tagging already tagged part 202004_297671_316874_302_339568. This i
s a bug., Stack trace (when copying this message, always include the lines below):... | 2020-05-01T06:56:21 | 2020-05-01T06:56:21 | {} | NONE |
ClickHouse | ClickHouse | 622,507,833 | 10,453 | alexey-milovidov | We found that odbc-bridge needs DateLUT for logging.
Let's close, because it doesn't affect anything.
| 2020-05-01T18:32:55 | 2020-05-01T18:32:55 | {} | MEMBER |
ClickHouse | ClickHouse | 622,555,569 | 10,585 | alexey-milovidov | Looks like there is some difference but my setup is noisy...
At least it is still slower than gcc. | 2020-05-01T20:31:56 | 2020-05-01T20:31:56 | {} | MEMBER |
ClickHouse | ClickHouse | 623,007,465 | 10,461 | wesm | You might be interested to take a closer look at the IO read coalescing work that has been happening for Parquet files lately specifically to address S3 use cases (note we also have bindings to the AWS C++ SDK in the core Arrow platform).
https://github.com/apache/arrow/pull/6744
I haven't looked at Clickhouse's ... | 2020-05-02T20:13:50 | 2020-05-02T20:13:50 | {} | NONE |
ClickHouse | ClickHouse | 623,012,384 | 10,461 | oandrew | There is another issue - aws s3 sdk reads the entire file into memory before returning it as an `std::istream`.
So for S3 + Parquet each file is loaded into memory twice.
https://github.com/ClickHouse-Extras/aws-sdk-cpp/blob/45dd8552d3c492defca79d2720bcc809e35654da/aws-cpp-sdk-core/source/http/curl/CurlHttpCli... | 2020-05-02T20:58:26 | 2020-05-02T20:58:26 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 623,034,873 | 10,630 | alexey-milovidov | It did not help. | 2020-05-03T00:40:33 | 2020-05-03T00:40:33 | {} | MEMBER |
ClickHouse | ClickHouse | 623,038,436 | 10,150 | alexey-milovidov | I have recreated toolchain for FreeBSD 11 and now it works perfectly. | 2020-05-03T01:21:53 | 2020-05-03T01:21:53 | {} | MEMBER |
ClickHouse | ClickHouse | 621,737,867 | 9,715 | filimonov | One more counter-intuitive usecase:
```
create table my_table (col_1 UInt64, col_2 UInt64, col_3 UInt64) Engine=MergeTree() ORDER BY tuple();
ALTER TABLE my_table UPDATE col_2 = col_1 + 10, col_3 = col_2 + 20 WHERE 1;
```
WIll it calculate col_3 based on new or old value of col_2?
```
insert into my_table(col... | 2020-04-30T10:04:27 | 2020-04-30T10:04:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,502,972 | 10,277 | alexey-milovidov | 1. `SELECT ignore(x) FROM (SELECT groupArray(number) AS x FROM numbers(33554433)) GROUP BY x FORMAT Null`
2. `INSERT INTO test.avro SELECT * FROM test.hits WHERE (intHash64(WatchID) % 100) = 0`
3. `SELECT sum(cityHash64(*)) FROM test.hits WHERE (intHash64(WatchID) % 100) = 0`
| 2020-05-01T18:20:51 | 2020-05-01T18:20:51 | {} | MEMBER |
ClickHouse | ClickHouse | 622,529,877 | 10,585 | alexey-milovidov | The size of
`clickhouse` was increased from 258 858 K to 276 878 K
`clickhouse-odbc-bridge` was lowered from 94 481 K to 86 246 K. | 2020-05-01T19:30:41 | 2020-05-01T19:30:41 | {} | MEMBER |
ClickHouse | ClickHouse | 621,690,357 | 10,176 | filimonov | Let's close. If will be needed - reopen, branch is here. | 2020-04-30T08:24:56 | 2020-04-30T08:24:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,714,291 | 10,563 | azat | > But the idea is OK and cast in case of different types is not important because otherwise, it will work as NOOP.
I'm playing with optimizing CAST away on ActionsVisitor.cpp but need to cleanup the code (to avoid even function call). But I'm not sure that I will finish it before stable release, since looks like it ... | 2020-04-30T09:14:11 | 2020-04-30T09:14:11 | {} | MEMBER |
ClickHouse | ClickHouse | 621,794,325 | 10,578 | vzakaznikov | @KochetovNicolai, this PR is ready for your review when you have time. The changes are quite small. | 2020-04-30T12:12:49 | 2020-04-30T12:12:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,803,944 | 10,594 | filimonov | Check `dmesg` on the server
Most probably OOMKiller killed the clickhouse | 2020-04-30T12:31:39 | 2020-04-30T12:31:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,982,016 | 10,555 | alesapin | Reopen if you will face the same error after update. | 2020-04-30T17:04:43 | 2020-04-30T17:04:43 | {} | MEMBER |
ClickHouse | ClickHouse | 621,783,261 | 10,593 | filimonov | Thanks, @4ertus2
So it relates to https://github.com/ClickHouse/ClickHouse/pull/8806 | 2020-04-30T11:47:43 | 2020-05-01T04:14:10 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,463,531 | 10,612 | den-crane | CH does not support correlated sub-queries.
you can use dictionaries
```
CREATE DICTIONARY item_dict ( eci_lu Int64, edu_lu Int64, pri_lu Int64, location String )
PRIMARY KEY eci_lu,edu_lu,pri_lu SOURCE(CLICKHOUSE(HOST 'localhost' PORT 9000
TABLE b DB 'default' USER 'default'))
LIFETIME(MIN 0 MAX 0) LAYOUT(... | 2020-05-01T16:40:52 | 2020-05-01T16:52:12 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,602,859 | 10,604 | vzakaznikov | > Still does not work.
@alexey-milovidov, clickhouse-server.log shows this exception. Do you think this is my bug?
```
2020.05.01 20:16:11.856206 [ 564 ] {d922d6cd-2bc4-4189-87a4-df64278e4a5d} <Error> executeQuery: Code: 32, e.displayText() = DB::Exception: Attempt to read after eof (version 20.5.1.3209) (from [... | 2020-05-01T23:00:48 | 2020-05-01T23:02:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,935,790 | 8,151 | amosbird | I just spent another couple of hours investigating this:
1. I fail to build nongnu libunwind with clickhouse
2. using gcc_eh doesn't crash but no stack symbols are produced
3. I created a minimal project depending on clickhouse's libunwind but failed to reproduce
4. I found that the problematic TU can be identifi... | 2020-05-02T11:02:38 | 2020-05-02T11:02:38 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 622,982,617 | 10,624 | shivakumarss | Thanks a lot @den-crane for the clarification. It helps.
It would really help people to have this feature in place otherwise the source csv needs to be modified or while importing i need to replace the / backslash with double quotes.
Closing this ticket now. | 2020-05-02T16:52:31 | 2020-05-02T16:52:31 | {} | NONE |
ClickHouse | ClickHouse | 622,989,012 | 10,625 | amosbird | hmm, even though that page has `<!DOCTYPE html>` but it is still opened as raw text | 2020-05-02T17:39:38 | 2020-05-02T17:39:38 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 621,778,962 | 10,593 | 4ertus2 | Bisect for 20.1 branch
```
commit 2334ae9e8799824a32d54220289d6988247f1f3d - FAIL
commit 2f232f082841cccd8126e777557eecc1b1110005 - build fail
commit 36b03f9c58e04b6fdd1836650e28015eaba794ed - OK
```
Broken here:
https://github.com/ClickHouse/ClickHouse/commit/2334ae9e8799824a32d54220289d6988247f1f3d
https://gi... | 2020-04-30T11:37:43 | 2020-04-30T11:44:51 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 621,810,148 | 10,594 | wutchzone | No sign of OOM killer in `dmesg`. There were about 100 GB free RAM. | 2020-04-30T12:44:07 | 2020-04-30T12:44:07 | {} | NONE |
ClickHouse | ClickHouse | 622,054,318 | 10,605 | den-crane | ```
cat /var/lib/clickhouse/user_files/rates.tsv
4990954156238030839 2018-12-31 21:00:00 2020-12-30 20:59:59 0.1 RU
CREATE DICTIONARY rates
(
hash_id UInt64,
start_date DateTime default '0000-00-00 00:00:00',
end_date DateTime default '0000-00-00 00:00:00',
price Float64,
currency String
)
PRIMARY KEY... | 2020-04-30T19:22:30 | 2020-04-30T19:22:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,383,696 | 10,607 | alexey-milovidov | Minor:
```
src/Processors/Formats/Impl/JSONAsStringRowInputFormat.cpp:127:16: error: redundant boolean literal in conditional return statement [readability-simplify-boolean-expr,-warnings-as-errors]
2020-04-30 21:54:36 return false;
2020-04-30 21:54:36 ~~~~~~~~~~~~~~~^~~~~~
``` | 2020-05-01T13:16:20 | 2020-05-01T13:16:20 | {} | MEMBER |
ClickHouse | ClickHouse | 622,661,500 | 7,019 | champtar | @filimonov you can add some more labels, this is still broken in latest stable/lts releases
```
$ podman run -d --network host --name clickhouse-server yandex/clickhouse-server:20.3.8.53
$ curl 'http://127.0.0.1:8123/?query=SELECT%201' -u 'default:' -m2
1
$ curl 'http://127.0.0.1:8123/?cancel_http_readonly_queries... | 2020-05-02T03:23:27 | 2020-05-02T03:23:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 622,940,385 | 10,585 | alexey-milovidov | Ok. And this PR is ready to merge. | 2020-05-02T11:40:42 | 2020-05-02T11:40:42 | {} | MEMBER |
ClickHouse | ClickHouse | 647,152,347 | 11,841 | den-crane | > Do you mean I should use daily partition? but on a daily basis, there are actually not that much of data.
I meant that your query processes excessive data because of `WHERE`
Probably it has sense to store a number of seconds from the date start (time w/o date) in a dedicated column and avoid a computation toInt64... | 2020-06-21T16:43:34 | 2020-06-21T16:43:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 647,160,637 | 11,836 | alexey-milovidov | Merged in #11843. | 2020-06-21T17:57:29 | 2020-06-21T17:57:29 | {} | MEMBER |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.