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 | 488,064,608 | 5,152 | janplus | @alexey-milovidov Thanks for your quick response. I'm on my vacation now, will check these errors after back to work. | 2019-04-30T18:33:01 | 2019-04-30T18:33:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,249,427 | 3,174 | blinkov | @q0987 do you have any further questions? | 2019-05-01T10:09:43 | 2019-05-01T10:09:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,335,611 | 5,160 | blinkov | @notbdu I've read the description again and as far as I can see it doesn't work only if you mention host3 (in `select count(*) from remote('host02,host03', system, "settings");`). Have you been doing this reconfiguration on all three hosts? (with clickhouse-server restarts)
Also all other diagnostics you provided me... | 2019-05-01T16:37:25 | 2019-05-01T16:37:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,044,879 | 5,003 | vitlibar | > About function signature, a proposal:
>
> In CH all geo related function get `x, y`, these new methods are accepting "lat, lon" but if we follow the same rule it should be `lon, lat` because `lon` is converted to `x` coordinate. Google Maps for some reason decided to swap the order but any serious GIS system uses ... | 2019-04-26T12:49:09 | 2019-04-26T12:51:51 | {} | MEMBER |
ClickHouse | ClickHouse | 487,123,652 | 5,127 | alexey-milovidov | Please also move a test from `bugs` directory. | 2019-04-26T16:45:15 | 2019-04-26T16:45:15 | {} | MEMBER |
ClickHouse | ClickHouse | 487,212,348 | 4,931 | vitlibar | > read_only is inconsistent with readonly
Both setting and constraint are named `readonly`. The comment message was incorrect. | 2019-04-26T21:50:21 | 2019-04-26T22:11:17 | {} | MEMBER |
ClickHouse | ClickHouse | 487,508,420 | 5,002 | ipoese | So... an update on the debug run.
We've moved the Production system to another environment and have been running the query pipeline that crashed the server for a few days now. But so far, we have not been able to re-produce the segfault with the 19.6.1 debug build. We've had multiple cashes of the query pipeline du... | 2019-04-29T09:07:29 | 2019-04-29T09:07:29 | {} | NONE |
ClickHouse | ClickHouse | 487,597,344 | 5,137 | den-crane | no force quote
```
| => cat aaa
1st,'2nd,3rd
| => cat aaa|clickhouse-client -q 'INSERT INTO crash_test FORMAT CSV'
Code: 27. DB::Exception: Cannot parse input: expected , at end of stream.: (at row 1)
```
force quote
```
| => cat aab
"1st","'2nd","3rd"
| => cat aab|clickhouse-client -q 'INSERT INTO cra... | 2019-04-29T14:16:21 | 2019-04-29T14:16:21 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,042,185 | 5,040 | danlark1 | > You can see this warning in `ClickHouse build check` for the latest commit (click to the clang-8 build and download logs):
> https://clickhouse-builds.s3.yandex.net/0/e914d0d67ae86fb2eb1536f670785ada2f58b836/build_log_416891604_1555767409.txt
There is no more warning after this commit https://github.com/yandex/Cl... | 2019-04-30T17:28:58 | 2019-04-30T17:28:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,062,715 | 5,152 | alexey-milovidov | Tests have failed. | 2019-04-30T18:27:55 | 2019-04-30T18:27:55 | {} | MEMBER |
ClickHouse | ClickHouse | 487,003,849 | 4,935 | zhang2014 | > We don't wait here https://github.com/yandex/ClickHouse/pull/4935/files#diff-8890e3b1de70b013b79201d37463a0d4R98. findFiles https://github.com/yandex/ClickHouse/blob/master/dbms/src/Storages/Distributed/DirectoryMonitor.cpp#L182 just looks at files in directory at current moment.
We will call `SYNC DISTRIBUTED` ... | 2019-04-26T10:04:43 | 2019-04-26T10:04:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,066,415 | 5,127 | abyss7 | Fixes #5066 | 2019-04-26T13:57:36 | 2019-04-26T13:57:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,256,582 | 5,125 | zombiemonkey | @pikhovkin. @den-crane is right. If you dig into the link there AresDB uses batch based processing for late data. Being creative - you could use the StorageJoin engine which allows for real-time in-memory updates/deduplication (you can also just run normal selects/queries from this and it is disk backed) and combine th... | 2019-04-27T05:26:43 | 2019-04-27T05:26:43 | {} | NONE |
ClickHouse | ClickHouse | 487,652,267 | 5,136 | alexey-milovidov | You can also add a performance test on `hits` dataset with URL and Referer columns.
(Just in case to prevent performance regressions in future.)
The test data can be downloaded:
```
sudo apt install wget xz-utils
wget https://clickhouse-datasets.s3.yandex.net/hits/tsv/hits_v1.tsv.xz
wget https://clickhouse-da... | 2019-04-29T16:38:17 | 2019-04-29T16:38:17 | {} | MEMBER |
ClickHouse | ClickHouse | 487,695,176 | 5,085 | kmatt | @abyss7 - Similar occurring in v19.5.3.8 via Yandex Docker container:
```
2019.04.29 18:27:06.754492 [ 1 ] {} <Error> Application: Caught exception while loading metadata: std::exception. Code: 1001, type: cppkafka::Exception, e.what() = Failed to create consumer handle: ssl.ca.location failed: ../contrib/ssl/crypt... | 2019-04-29T18:38:54 | 2019-04-29T19:05:39 | {} | NONE |
ClickHouse | ClickHouse | 488,024,787 | 5,040 | danlark1 | > ```
> ../contrib/lz4/lib/lz4.c:102:9: warning: 'LZ4_DISABLE_DEPRECATE_WARNINGS' macro redefined [-Wmacro-redefined]
> #define LZ4_DISABLE_DEPRECATE_WARNINGS /* due to LZ4_decompress_safe_withPrefix64k */
> ```
With latest commit, this message should disappear
https://github.com/lz4/lz4/blob/7a4e3b1fac5cd9d4ec7... | 2019-04-30T16:38:53 | 2019-04-30T16:38:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,039,465 | 5,040 | alexey-milovidov | You can see this warning in `ClickHouse build check` for the latest commit (click to the clang-8 build and download logs):
https://clickhouse-builds.s3.yandex.net/0/e914d0d67ae86fb2eb1536f670785ada2f58b836/build_log_416891604_1555767409.txt | 2019-04-30T17:20:49 | 2019-04-30T17:20:49 | {} | MEMBER |
ClickHouse | ClickHouse | 488,205,560 | 5,160 | notbdu | IPV4 is working fine.
```
host01:~$ curl -4 -vvv host02:9001
* Rebuilt URL to: host02:9001/
* Hostname was NOT found in DNS cache
* Trying $HOST02IP...
* Connected to host02 ($HOST02IP) port 9001 (#0)
> GET / HTTP/1.1
> User-Agent: curl/7.38.0
> Host: host02:9001
> Accept: */*
>
* HTTP 1.0, assume close ... | 2019-05-01T03:56:42 | 2019-05-01T03:56:42 | {} | NONE |
ClickHouse | ClickHouse | 486,964,359 | 4,935 | alesapin | > `SYNC DISTRIBUTED` is more convenient for writing tests, as well as some special scenarios, such as ensuring that no distributed data needs to be sent before data migration.
We don't wait here https://github.com/yandex/ClickHouse/pull/4935/files#diff-8890e3b1de70b013b79201d37463a0d4R98. `findFiles` https://github.... | 2019-04-26T07:55:37 | 2019-04-26T07:55:37 | {} | MEMBER |
ClickHouse | ClickHouse | 486,977,134 | 5,115 | blinkov | @yjhmelody @chaelub could you tell about your usecases in more detail?
why your client have direct access to server files in the first place? | 2019-04-26T08:38:56 | 2019-04-26T08:38:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,068,936 | 5,115 | chaelub | just checked the logs. So it seems like there is some casting troubles.
`2019.04.26 17:01:07.446018 [ 24 ] {b4a9165c-a36e-484b-868e-8f439aca3403} <Error> executeQuery: Code: 368, e.displayText() = DB::Exception: Bad cast from type DB::ColumnVector<unsigned char> to DB::ColumnVector<long> (version 19.5.3.8 (official ... | 2019-04-26T14:04:52 | 2019-04-26T14:14:27 | {} | NONE |
ClickHouse | ClickHouse | 487,316,998 | 4,639 | ivan-v-kush | @alexey-milovidov
I think, the problem is that in this PR data part's mutex is locked in threadpool's threads and unlocked in the main thread
What should be done? Add supression in ThreadSanitizer? Or refactor the creating/commiting of `AlterDataPartTransactionPtr`s?
**Description:**
Tsan reports:
```
WAR... | 2019-04-27T20:25:33 | 2019-04-27T21:18:40 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,351,891 | 5,132 | blinkov | There are such plans, but they are not short term. Meanwhile some of window functions can be emulated otherwise. See #1469 | 2019-04-28T07:11:02 | 2019-04-28T07:11:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,757,235 | 5,148 | alexey-milovidov | The reason is that `max_block_size` has no effect on `StripeLog` and `Memory` engine. Data in `StripeLog` and `Memory` is stored in blocks. And data is read in the same blocks as was stored in a table, regardless to `max_block_size`.
In contrast, in `Log` and `TinyLog`, data is stored in almost contiguous columns th... | 2019-04-29T21:52:00 | 2019-04-29T21:52:00 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 487,757,677 | 5,148 | alexey-milovidov | To get atomicity for all block sizes:
```
SET
max_insert_block_size = 10000000000,
min_insert_block_size_rows = 10000000000,
min_insert_block_size_bytes = 10000000000000;
```
It will cost high memory usage for large INSERTs. | 2019-04-29T21:53:39 | 2019-04-29T21:54:19 | {} | MEMBER |
ClickHouse | ClickHouse | 488,348,074 | 5,031 | alexey-milovidov | Fixed in master. | 2019-05-01T17:20:03 | 2019-05-01T17:20:03 | {} | MEMBER |
ClickHouse | ClickHouse | 487,349,594 | 5,003 | Enmk | > > About function signature, a proposal:
> > In CH all geo related function get `x, y`, these new methods are accepting "lat, lon" but if we follow the same rule it should be `lon, lat` because `lon` is converted to `x` coordinate. Google Maps for some reason decided to swap the order but any serious GIS system uses ... | 2019-04-28T06:36:35 | 2019-04-28T06:36:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,649,143 | 5,135 | alexey-milovidov | Ok. | 2019-04-29T16:29:26 | 2019-04-29T16:29:26 | {} | MEMBER |
ClickHouse | ClickHouse | 488,096,016 | 5,153 | ghost | @blinkov
* Which hardware have you used?
- I have 30 machines (dual intel silver 4110, 256GB Memory, 24TB HDD RAID 1+0)
* How large was the batches?
- Batches insert 3,000,000,000 rows into the ClickHouse. Each row has 200 columns and the size of a row ranges from 200 to 800 bytes.
* How much faster do you n... | 2019-04-30T20:06:32 | 2019-04-30T20:10:50 | {} | NONE |
ClickHouse | ClickHouse | 487,720,770 | 5,085 | kmatt | So I could turn up logging to DEBUG level, I stopped all four Docker containers, and changed the logging level with the new log path. Upon restart of the containers, none of the 4 servers would accept clickhouse-client connections, and Kafka ingest rate was 0.
Only way to get them started was again to drop kafka tab... | 2019-04-29T19:56:54 | 2019-04-29T19:56:54 | {} | NONE |
ClickHouse | ClickHouse | 487,946,385 | 5,137 | den-crane | CSV parser works as expected
https://clickhouse.yandex/docs/en/interfaces/formats/#csv
>Both double and single quotes are supported.
Check TSV https://clickhouse.yandex/docs/en/interfaces/formats/#tabseparated | 2019-04-30T13:12:12 | 2019-04-30T13:12:12 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,118,070 | 5,153 | den-crane | >24TB HDD RAID 1+0
And how many disks? You need as many disks as possible for random i/o because each insert creates 400 files (200cols*2).
Though 30 machines should easily ingest 1,000,000 rows/second.
Try bigger batches / upload with clickhouse-client from TSV files by 1mil rows directly to Mergetree table (wi... | 2019-04-30T21:14:07 | 2019-04-30T21:14:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,322,342 | 4,282 | kmatt | Seeing this behavior again on 19.1.4 and 19.5.3.8 | 2019-05-01T15:52:17 | 2019-05-01T15:52:17 | {} | NONE |
ClickHouse | ClickHouse | 488,372,181 | 5,160 | notbdu | Oddly, queries work between `host01`<-> `host03` but do not work from `host02` <-> `host01` or `host02` <-> `host03`. I'll document some of the debugging results I've seen below.
Running `strace` on `host02` and `host03` both show similar results as they correctly bind to the configured port.
```
host03:~$ sudo -... | 2019-05-01T18:38:59 | 2019-05-01T18:42:06 | {} | NONE |
ClickHouse | ClickHouse | 486,942,563 | 5,095 | YiuRULE | Thanks for the suggestion, now both function return a tuple with an IPv4 and IPv6! | 2019-04-26T06:31:38 | 2019-04-26T06:31:38 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,002,422 | 5,119 | nicelulu | https://github.com/yandex/ClickHouse/issues/5108 | 2019-04-26T09:59:48 | 2019-04-26T09:59:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,061,948 | 5,125 | den-crane | ```
create table xx (A String, X UInt64) engine=ReplacingMergeTree order by A;
insert into xx values ('a', 1);
insert into xx values ('a', 2);
select * from xx final;
┌─A─┬─X─┐
│ a │ 2 │
└───┴───┘
optimize table xx final;
select count() from xx
┌─count()─┐
│ 1 │
└─────────┘
``` | 2019-04-26T13:44:19 | 2019-04-26T13:52:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,073,538 | 5,125 | den-crane | >This is no nothing different by speed from ordinary relations databases
¯\_(ツ)_/¯
You need another realm then, with real magic. | 2019-04-26T14:18:06 | 2019-04-26T14:18:06 | {
"eyes": 1,
"laugh": 10
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,185,691 | 4,931 | alexey-milovidov | read_only is inconsistent with readonly | 2019-04-26T20:09:40 | 2019-04-26T20:09:40 | {} | MEMBER |
ClickHouse | ClickHouse | 487,533,960 | 5,126 | abyss7 | Yes, it's definitely the bug. Will find out which versions it affects. | 2019-04-29T10:44:21 | 2019-04-29T10:44:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,002,592 | 5,153 | blinkov | @achimbab how fast were inserts that you have tried, which hardware have you used, how large was the batches and how much faster do you need it to be? | 2019-04-30T15:39:19 | 2019-04-30T15:45:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,204,682 | 751 | abraithwaite | The code here suggests that it shouldn't be possible to `GROUP BY` in a select query where the fields in the group by aren't included in the original primary key for the materialized view:
https://github.com/yandex/ClickHouse/blob/83a35d7644cad823533c3a6459c8c777f9d10481/dbms/src/DataStreams/AggregatingSortedBlockIn... | 2019-05-01T03:46:15 | 2019-05-01T03:46:15 | {} | NONE |
ClickHouse | ClickHouse | 488,213,987 | 5,160 | blinkov | @notbdu what do you have in your config.xml around here:
https://github.com/yandex/ClickHouse/blob/master/dbms/programs/server/config.xml#L70-L72
You could try to keep both lines or change one to another.
Also could you show the output of `iptables -L`? | 2019-05-01T05:30:58 | 2019-05-01T08:42:46 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,596,393 | 5,137 | den-crane | try force quote
COPY (select * from crash_test) TO PROGRAM 'clickhouse-client --query="INSERT INTO crash_test FORMAT CSV"' with csv force quote *;
| 2019-04-29T14:13:52 | 2019-04-29T14:13:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,630,370 | 5,125 | den-crane | > It sounds very cool. When I loaded data in CH (40kk) it didnt collapse the duplicates while ~10-12 hours. Duplicates was collapsed only after `optimize ...` command.
Merges are eventual. They may never happen because of parts size for example. | 2019-04-29T15:38:46 | 2019-04-29T15:39:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,651,309 | 5,112 | blinkov | сс @alex-krash | 2019-04-29T16:35:34 | 2019-04-29T16:35:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,747,909 | 2,198 | alexey-milovidov | Implemented in #5138.
Merged into master. | 2019-04-29T21:19:05 | 2019-04-29T21:19:05 | {} | MEMBER |
ClickHouse | ClickHouse | 488,338,201 | 5,161 | alexey-milovidov | With four letter commands:
- `srvr` or `stat` - check that only one node is leader; check that number of nodes does not differ more than 1% and more than 1000.
- `conf` - check that every `serverId` is different;
| 2019-05-01T16:46:16 | 2019-05-01T16:46:16 | {} | MEMBER |
ClickHouse | ClickHouse | 488,359,348 | 5,160 | blinkov | What about host1 -> host3 and host3 -> host2? | 2019-05-01T17:57:22 | 2019-05-01T17:57:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 486,991,832 | 5,115 | yjhmelody | @blinkov hi! because of our powerful server and poor net , it's better let data store in server. Also, so many datum shared by server. If client can set a path to server's data, sometime very convenient.
Any way, I have solved my problem in some way. | 2019-04-26T09:27:24 | 2019-04-26T09:27:24 | {} | NONE |
ClickHouse | ClickHouse | 487,020,567 | 5,030 | YiuRULE | Thanks, I applied the change.
It seem we got a weird bug with the CI process when tried to build using g++. I tried to recompile from scratch, both on this branch, but also on the latest master branch and I got this
```
[7/87] cd /home/guillaume/ClickHouse &...d/dbms/programs/clang/headers/19.7.1.1
Use one of ... | 2019-04-26T11:11:10 | 2019-04-26T11:11:50 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,078,204 | 5,125 | pikhovkin | :-)
> [When AresDB receives an upsert batch for ingestion, it first writes the upsert batch to redo logs for recovery. After an upsert batch is appended to the end of the redo log, AresDB then identifies and skips late records on fact tables for ingestion into the live store. A record is considered “late” if its eve... | 2019-04-26T14:30:53 | 2019-04-26T14:30:53 | {} | NONE |
ClickHouse | ClickHouse | 487,332,474 | 5,115 | laoduo | #数据导出
clickhouse-client --host="host25080" --max_threads="10" --format_csv_delimiter="|" --query="select * from ares.logentity_test2_all FORMAT CSV" > logentity_test2_all.csv
#数据导入
clickhouse-client --host="host25080" --format_csv_delimiter="|" --query="insert into ares.logentity_20181128 FORMAT CSV" </data/clickh... | 2019-04-28T01:32:27 | 2019-04-28T01:32:27 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 487,522,758 | 5,134 | zhang2014 | Original article -> 来源文章 | 2019-04-29T09:59:12 | 2019-04-29T09:59:12 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,617,942 | 5,137 | den-crane | ah, another option by CH --format_csv_allow_single_quotes
cat aaa|clickhouse-client --format_csv_allow_single_quotes 0 -q 'INSERT INTO crash_test FORMAT CSV' | 2019-04-29T15:07:37 | 2019-04-29T15:07:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,027,301 | 5,040 | alexey-milovidov | I have built on master with all updated submodules about now and got this message. | 2019-04-30T16:46:08 | 2019-04-30T16:46:08 | {} | MEMBER |
ClickHouse | ClickHouse | 488,146,083 | 5,160 | notbdu | Inspecting the packet capture, it looks like the destination IP from the client -> server is an IPV4 address.
Also looks like the client -> server send/receive packets from each other attempting to set up a TCP connection but fail to do so.
OTOH, I have verified that I can establish a TCP connection in both direc... | 2019-04-30T23:03:18 | 2019-04-30T23:03:18 | {} | NONE |
ClickHouse | ClickHouse | 488,329,706 | 5,160 | notbdu | I've tried both of the following configurations.
```
<!-- <listen_host>::</listen_host> -->
<!-- Same for hosts with disabled ipv6: -->
<listen_host>0.0.0.0</listen_host>
```
and
```
<listen_host>::</listen_host>
<!-- Same for hosts with disabled ipv6: -->
<!-- <listen_host>0.0.0.0</listen_host> -->
```
and... | 2019-05-01T16:17:15 | 2019-05-01T16:18:38 | {} | NONE |
ClickHouse | ClickHouse | 487,036,353 | 5,111 | filimonov | Closed in favor to #5121 | 2019-04-26T12:16:52 | 2019-04-26T12:16:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,065,606 | 5,125 | pikhovkin | I need guarantees of deduplication in realtime.
*MergeTree does not guarantee deduplication.
[`Data deduplication occurs only during a merge. Merging occurs in the background at an unknown time, so you can't plan for it. Some of the data may remain unprocessed. `](https://clickhouse.yandex/docs/en/operations/table... | 2019-04-26T13:55:12 | 2019-04-26T13:55:12 | {
"+1": 4
} | NONE |
ClickHouse | ClickHouse | 487,584,823 | 5,134 | neverlee | 感觉 ”Original article“ 翻译成”原文“ 好些 | 2019-04-29T13:43:46 | 2019-04-29T13:43:46 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,017,799 | 5,040 | alexey-milovidov | ```
../contrib/lz4/lib/lz4.c:102:9: warning: 'LZ4_DISABLE_DEPRECATE_WARNINGS' macro redefined [-Wmacro-redefined]
#define LZ4_DISABLE_DEPRECATE_WARNINGS /* due to LZ4_decompress_safe_withPrefix64k */
``` | 2019-04-30T16:18:51 | 2019-04-30T16:18:51 | {} | MEMBER |
ClickHouse | ClickHouse | 488,340,641 | 5,160 | notbdu | It doesn't work in the strictly `host01` -> `host02` case.
Hmmm, it looks like `nc` is able to establish a TCP connection when CH is running and listening on port 9000?
```
host01:~$ nc -vz 10.10.1.111 9000
nc: host02 (10.10.1.111) 9000 [9000] open
host01:~$ nc -4 -vz 10.10.1.111 9000
nc: host02 (10.10.1.111... | 2019-05-01T16:54:21 | 2019-05-01T16:54:21 | {} | NONE |
ClickHouse | ClickHouse | 487,374,212 | 2,198 | alexey-milovidov | It is assigned to @CurtizJ | 2019-04-28T12:22:43 | 2019-04-28T12:22:43 | {} | MEMBER |
ClickHouse | ClickHouse | 487,599,751 | 5,125 | zombiemonkey | Really? I just managed 5 million updates on a 'current day' table ala AresDB in < 1 second. Haven't tested multiple inserts/single inserts but these can always be dealt with using buffered inserts or buffer tables. You could also chain the updates together behind a materialized view if required. You will need some orch... | 2019-04-29T14:22:40 | 2019-04-29T14:22:40 | {
"+1": 1,
"-1": 1,
"heart": 6
} | NONE |
ClickHouse | ClickHouse | 487,899,043 | 5,137 | artem-pershin | > ah, another option by CH --format_csv_allow_single_quotes
>
> cat aaa|clickhouse-client --format_csv_allow_single_quotes 0 -q 'INSERT INTO crash_test FORMAT CSV'
thank you, both answers fix the issue. So it's not a bug. But im not sure if other symbols can also crash parser if we escape apostrophe. | 2019-04-30T10:16:08 | 2019-04-30T10:16:08 | {} | NONE |
ClickHouse | ClickHouse | 488,083,312 | 1,821 | xichen2020 | +1, running into similar issues when adding/replacing nodes during our tests, any ETA on the fix? | 2019-04-30T19:27:17 | 2019-04-30T19:27:17 | {
"eyes": 1
} | NONE |
ClickHouse | ClickHouse | 488,113,443 | 5,160 | den-crane | Seems clickhouse-client uses IPV6
>executeQuery: (from [::ffff:127.0.0.1]:60802) select count(*) from remote('host02,host03'
>executeQuery: (from [::ffff:10.00.000.100]:46260
And clickhouse-server uses IPV4
>Timeout: connect timed out: 10.10.1.111:9000
And as I see from tcpdump you have an network issue wi... | 2019-04-30T20:58:57 | 2019-04-30T21:04:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,382,991 | 5,160 | blinkov | > Oddly, queries work between host01<-> host03 but do not work from host02 <-> host01 or host02 <-> host03. I'll document some of the debugging results I've seen below.
@notbdu then likely there's an easy solution: wipe host2 clean and replace it with fresh server, then let replication do the rest (or do backup+rest... | 2019-05-01T19:08:46 | 2019-05-01T19:08:46 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,124,053 | 5,119 | alexey-milovidov | Please add description for changelog. | 2019-04-26T16:46:31 | 2019-04-26T16:46:31 | {} | MEMBER |
ClickHouse | ClickHouse | 487,417,260 | 5,133 | alexey-milovidov | Ok.
But
```
&& munmap(largeBlock, sz) == 0
```
looks weird. | 2019-04-28T21:31:25 | 2019-04-28T21:31:25 | {} | MEMBER |
ClickHouse | ClickHouse | 487,434,533 | 5,119 | alexey-milovidov | Now everything is perfect, thank you! | 2019-04-29T01:51:31 | 2019-04-29T01:51:31 | {} | MEMBER |
ClickHouse | ClickHouse | 487,458,074 | 5,134 | blinkov | cc @zhang2014 @sundy-li | 2019-04-29T05:52:16 | 2019-04-29T05:52:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,046,684 | 5,040 | alexey-milovidov | Ok. | 2019-04-30T17:42:15 | 2019-04-30T17:42:15 | {} | MEMBER |
ClickHouse | ClickHouse | 486,975,844 | 5,115 | chaelub | Hello. I have a similar problem.
I tried this way (by analogy with insert from csv):
`cat file_name.bin | clickhouse-client --password=my_password --database=some_db --query="INSERT INTO some_table FORMAT Native"`
but I got an error:
`Code: 210. DB::NetException: Connection reset by peer, while writing to socket (1... | 2019-04-26T08:34:37 | 2019-04-26T08:34:37 | {} | NONE |
ClickHouse | ClickHouse | 487,008,167 | 5,119 | nicelulu | Performance is slightly worse. Or we need to declare that a row of data cannot satisfy both the first and NTH conditions. | 2019-04-26T10:20:37 | 2019-04-26T10:24:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,081,538 | 5,088 | abyss7 | > @palasonic1 will do this task.
Maybe reference and close this one? | 2019-04-26T14:40:10 | 2019-04-26T14:40:10 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,087,623 | 5,125 | den-crane | AresDB is in-memory database.
Do you want the size of your data be equal of your's server RAM as well?
Of course everybody dreams about this feature, but How? Disks are too slow for real-time merge within ingestion. | 2019-04-26T14:57:18 | 2019-04-26T15:18:10 | {
"laugh": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,552,935 | 5,125 | pikhovkin | @den-crane I talk about the idea for fast (more or less fast) updating of data.
Now the speed of updating of data in CH is ~8qps (only updates) and ~16qps (1/3 updates and 2/3 inserts). This is through `ALTER ... UPDATE ...` mechanism for each row. Аnd after 100к-200к of these updates, the memory starts to leak.
I n... | 2019-04-29T12:03:50 | 2019-04-29T12:05:18 | {} | NONE |
ClickHouse | ClickHouse | 487,575,477 | 5,125 | zombiemonkey | @pikhovkin. It would be good to understand your full requirements/data volumes. Basically you're describing an OLTP or key/value store and not an OLAP store.
I noted a way to achieve the same functionality as AresDB above by using StorageJoin. This is an in-memory table that syncs to disk. You can update records by ... | 2019-04-29T13:16:52 | 2019-04-29T13:16:52 | {} | NONE |
ClickHouse | ClickHouse | 487,659,372 | 5,108 | abyss7 | Looks like this was fixed. | 2019-04-29T16:57:50 | 2019-04-29T16:57:50 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,023,017 | 5,070 | danlark1 | For all processors, except ARM and PowerPC we got 2-3% performance gain of decompression (including AMD chips).
For ARM and PPC bayesian bandits algorithm doesn't choose the 32 byte version almost never -- possibly because of loop unrolling but need to investigate more. Slowdown is 0.1-0.2% because of some additiona... | 2019-04-30T16:33:45 | 2019-04-30T16:33:45 | {
"+1": 3
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,442,390 | 5,116 | alexey-milovidov | It does not work:
```
2019.04.29 05:28:59.704779 [ 872 ] {} <Error> BaseDaemon: ########################################
2019.04.29 05:28:59.704981 [ 872 ] {} <Error> BaseDaemon: (version 19.7.1.404) (from thread 34) Received signal Segmentation fault (11).
2019.04.29 05:28:59.705097 [ 872 ] {} <Error> BaseDaemon: ... | 2019-05-01T21:48:13 | 2019-05-01T21:48:13 | {} | MEMBER |
ClickHouse | ClickHouse | 486,979,618 | 5,115 | chaelub | @blinkov In my case there isn't communications between client and server, it's all about the clickhouse-client tool on the same machine where clickhouse-server installed.
Could you explain how to insert data from Native format to a CH table? | 2019-04-26T08:47:25 | 2019-04-26T08:47:40 | {} | NONE |
ClickHouse | ClickHouse | 487,067,695 | 5,125 | pikhovkin | > select * from xx final;
This is no nothing different by speed from ordinary relations databases
> optimize table xx final;
It's bulllshit for realtime | 2019-04-26T14:01:17 | 2019-04-26T14:01:17 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 487,649,681 | 5,135 | alexey-milovidov | PS. https://github.com/yandex/ClickHouse/commit/2cf0ab341ecec546022a61f9706210221d135cb9 https://github.com/yandex/ClickHouse/commit/3c8dce62a4d6553e181ae3bd9b00fe441ddac062 https://github.com/yandex/ClickHouse/commit/f999100e562672e7be0d890e3b1897514fedd7ed | 2019-04-29T16:30:56 | 2019-04-29T16:32:25 | {} | MEMBER |
ClickHouse | ClickHouse | 487,766,140 | 5,031 | alex-zaitsev | Easily reproducible test:
```
create table test_in (a LowCardinality(String)) Engine = MergeTree order by a;
insert into test_in values ('a');
select * from test_in where a in ('a');
Received exception from server (version 19.5.3):
Code: 49. DB::Exception: Received from localhost:9000, 127.0.0.1. DB::Exce... | 2019-04-29T22:28:05 | 2019-04-29T22:28:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,018,349 | 5,013 | alexey-milovidov | The library was updated.
Mixed with our implementation of decompression loop, it became even faster. | 2019-04-30T16:20:18 | 2019-04-30T16:20:18 | {
"hooray": 2
} | MEMBER |
ClickHouse | ClickHouse | 488,323,024 | 5,153 | blinkov | @achimbab how exactly do you perform inserts and how many rows/second you see at the moment?
Do you insert into `Distributed` table or directly to shards? How MergeTree tables are configured? | 2019-05-01T15:54:34 | 2019-05-01T15:54:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 486,963,543 | 5,107 | blinkov | @AlexZFX it sounds like some symlink issue: try to remove this file, commit and then create the new one in it's place. | 2019-04-26T07:52:47 | 2019-04-26T07:52:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,027,811 | 5,040 | danlark1 | Even with this commit?
https://github.com/yandex/ClickHouse/commit/3f1523a80a0a19c301e3d31fbb59948c9130c069 | 2019-04-30T16:47:37 | 2019-04-30T16:47:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 488,147,496 | 5,160 | den-crane | > Inspecting the packet capture, it looks like the destination IP from the client -> server is an IPV4 address.
Yes. And the problem your IPV4 is not working.
>Also looks like the client -> server send/receive packets from each other attempting to set up a TCP connection but fail to do so.
Yes. Because your IP... | 2019-04-30T23:09:52 | 2019-04-30T23:12:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,122,854 | 5,088 | alexey-milovidov | Ok. | 2019-04-26T16:42:32 | 2019-04-26T16:42:32 | {} | MEMBER |
ClickHouse | ClickHouse | 487,256,871 | 5,125 | zombiemonkey | Check out the example of StorageJoin here. The docs probably need to be updated and don't provide the full context of what is possible with this engine.
https://github.com/yandex/ClickHouse/blob/master/dbms/tests/queries/0_stateless/00800_versatile_storage_join.sql | 2019-04-27T05:33:51 | 2019-04-27T05:33:51 | {} | NONE |
ClickHouse | ClickHouse | 487,341,771 | 5,119 | nicelulu | > Please add description for changelog.
Do I need to add code comments? | 2019-04-28T04:13:46 | 2019-04-28T04:13:46 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,586,242 | 5,125 | pikhovkin | It sounds very cool. When I loaded data in CH (40kk) it didnt collapse the duplicates while ~10-12 hours. Duplicates was collapsed only after `optimize ...` command. | 2019-04-29T13:47:37 | 2019-04-29T13:47:37 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 487,629,767 | 5,125 | den-crane | >I tried Vertica, it has MERGE mechanism for updates-inserts.
>The speed for merge was ~55-70qps (only updates). And Vertica works very stable.
@pikhovkin
Did you try it with 2B rows table?
~55-70qps of MERGE INTO ? Or batched update by one MERGE INTO ?
| 2019-04-29T15:37:15 | 2019-04-29T15:37:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,720,246 | 5,147 | champtar | Why not just put a different group name ? | 2019-04-29T19:55:22 | 2019-04-29T19:55:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 487,964,804 | 5,125 | pikhovkin | @zombiemonkey I need to update the data periodically (every 5-15 minutes) for 3 days. There are text fields in the data.
@den-crane I load 40kk rows by 1k rows. Rows contains ~100 columns. | 2019-04-30T14:03:34 | 2019-04-30T14:03:34 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 488,336,624 | 5,160 | blinkov | > OTOH, I have verified that I can establish a TCP connection in both directions by running a simple web server on port 9000.
Can you do that with `nc -vz 10.10.1.111 9000` from host1 (while there's a ClickHouse on other side)? | 2019-05-01T16:40:56 | 2019-05-01T16:40:56 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.