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 | 403,138,087 | 2,111 | silviucpp | OK I understand,
But there is any workaround I can use to avoid errors for the examples I provided ? | 2018-07-06T20:26:02 | 2018-07-06T20:26:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,335,233 | 2,383 | alexey-milovidov | It can be easily solved with DDLGuard. | 2018-07-09T01:30:40 | 2018-07-09T01:30:40 | {} | MEMBER |
ClickHouse | ClickHouse | 403,960,062 | 2,627 | alexey-milovidov | Totally wrong. | 2018-07-10T20:46:03 | 2018-07-10T20:46:03 | {} | MEMBER |
ClickHouse | ClickHouse | 403,147,487 | 2,111 | alexey-milovidov | ```select CAST([toInt64(-9223372036854775808), toInt64(9223372036854775807)] AS Array(Nullable(Int64)))``` | 2018-07-06T21:08:57 | 2018-07-06T21:08:57 | {} | MEMBER |
ClickHouse | ClickHouse | 403,246,580 | 1,770 | simPod | Hey, did you solve it? How to filter with uuid in WHERE? | 2018-07-07T22:22:32 | 2018-07-07T22:22:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,625,827 | 2,617 | alexey-milovidov | We run limited amount of tests with a limited version of ClickHouse in Travis:
https://travis-ci.org/yandex/ClickHouse/jobs/401943668#L7772 | 2018-07-09T21:26:44 | 2018-07-09T21:26:52 | {} | MEMBER |
ClickHouse | ClickHouse | 403,879,790 | 2,447 | alexey-milovidov | Yes, it requires multiple changes. I will address them later. | 2018-07-10T16:12:09 | 2018-07-10T16:12:09 | {} | MEMBER |
ClickHouse | ClickHouse | 404,269,707 | 1,322 | alexey-milovidov | The most simple case:
```
SELECT CAST(arrayJoin(['1', NULL]), 'Nullable(UInt8)')
Received exception from server (version 1.1.54391):
Code: 32. DB::Exception: Received from localhost:9000, 127.0.0.1. DB::Exception: Attempt to read after eof: Cannot parse UInt8 from String, because value is too short.
``` | 2018-07-11T18:38:18 | 2018-07-11T18:38:18 | {} | MEMBER |
ClickHouse | ClickHouse | 403,251,634 | 1,546 | alexey-milovidov | I cannot reproduce this issue in master.
A few notes: `VALUES(now(), now(), 1, 1);`
- I have to wait one second between INSERTS, otherwise `now()` is the same and rows will be deduplicated;
- First `now()` is converted to `Date` as number - better to replace it with `today()` (but it doesn't alter results).
| 2018-07-08T00:16:16 | 2018-07-08T00:16:16 | {} | MEMBER |
ClickHouse | ClickHouse | 403,257,387 | 2,281 | alexey-milovidov | CPU time is spent in `statfs` system call, that is used to check free space before assigning merge
and in `readdir` for `DB::MergeTreeData::clearOldTemporaryDirectories`, `DB::MergeTreeData::clearOldPartsFromFilesystem`.
Also some time is spent in `DB::Context::getPartLog` for getting configuration value. | 2018-07-08T02:37:59 | 2018-07-08T02:37:59 | {} | MEMBER |
ClickHouse | ClickHouse | 403,259,656 | 2,281 | alexey-milovidov | When I do test:
```
seq 1 100000 | sed -r -e 's/^.+$/INSERT INTO multiple_mergetrees.table\0 SELECT rand()/' | clickhouse-benchmark -c16
```
I see periodic stalls for 10s of seconds. | 2018-07-08T03:34:45 | 2018-07-08T03:34:45 | {} | MEMBER |
ClickHouse | ClickHouse | 403,282,319 | 1,546 | silviucpp | Hello,
This was a pretty old report. BackgroundSchedulePool was not yet available at that time. Anyway I'll try again to see if I can or not on master and let you know. Maybe was fixed in the meantime. | 2018-07-08T11:52:15 | 2018-07-08T11:52:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,625,198 | 2,434 | alexey-milovidov | Yes, `treat_local_as_remote` is not configurable right now. We have plans to make it available through settings. | 2018-07-09T21:24:14 | 2018-07-09T21:24:14 | {} | MEMBER |
ClickHouse | ClickHouse | 403,822,175 | 2,596 | soulne4ny | > Why are you interested in absolute values of that?
The goal was to combine time part of one DateTime value with date part of another. It was while trying different ways to generate time values in between two moments in time. Final solution contained dates without time. Dates are capable of ±1 and it is enouth.
Ho... | 2018-07-10T13:29:30 | 2018-07-10T13:29:30 | {} | NONE |
ClickHouse | ClickHouse | 403,865,522 | 2,447 | Tezd | @alexey-milovidov Are there other changes that are needed ? | 2018-07-10T15:30:40 | 2018-07-10T15:30:40 | {} | NONE |
ClickHouse | ClickHouse | 404,510,635 | 1,939 | PangKuo | @fr05t1k any way to use customize user for replicated tables? I have the same problem, when I tried to drop partition on non-leader replica , I got an error | 2018-07-12T13:24:02 | 2018-07-12T13:24:02 | {} | NONE |
ClickHouse | ClickHouse | 404,716,880 | 2,637 | alexey-milovidov | Need to add an example in comments in `users.xml` file. | 2018-07-13T03:42:31 | 2018-07-13T03:42:31 | {} | MEMBER |
ClickHouse | ClickHouse | 403,123,969 | 2,115 | alexey-milovidov | The code is just plain wrong (no magic). | 2018-07-06T19:18:40 | 2018-07-06T19:18:40 | {} | MEMBER |
ClickHouse | ClickHouse | 403,270,832 | 2,608 | arduanov | gcc-8
```
-DCMAKE_C_FLAGS_RELEASE=-DNDEBUG
-DCMAKE_CXX_FLAGS_RELEASE=-DNDEBUG
-DCMAKE_INSTALL_PREFIX=/usr/local/Cellar/clickhouse/1.1.54388
-DCMAKE_BUILD_TYPE=Release
-DCMAKE_FIND_FRAMEWORK=LAST
-DCMAKE_VERBOSE_MAKEFILE=ON
-Wno-dev
-DENABLE_ICU=0,
-DENABLE_TESTS=0,
-DENABLE_TCMALLOC=0,
-DUSE_INTERNAL_BOOST_... | 2018-07-08T08:10:52 | 2018-07-08T08:10:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,222,748 | 2,612 | filimonov | 1) very often table can be redesigned to avoid JOINs in Clickhouse (using Arrays / external Dictionaries or denormalization).
2) you can just have a table and put data there on some regular basis like that:
```INSERT INTO ... SELECT ... WHERE ... ``` | 2018-07-11T16:00:04 | 2018-07-11T16:00:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,119,062 | 2,115 | alexey-milovidov | Yes, it's a bug. I'll try to investigate. | 2018-07-06T18:56:07 | 2018-07-06T18:56:07 | {} | MEMBER |
ClickHouse | ClickHouse | 403,120,462 | 2,115 | alexey-milovidov | > item_arg->onlyNull() should evaluate to true and not false for this case
It's for the case when you write `arrayHas([...], NULL)` | 2018-07-06T19:01:59 | 2018-07-06T19:01:59 | {} | MEMBER |
ClickHouse | ClickHouse | 403,273,815 | 2,605 | alex-zaitsev | What about SQL domains? That does not change internal representation, mostly type name aliasing.
https://www.postgresql.org/docs/9.1/static/sql-createdomain.html | 2018-07-08T09:08:24 | 2018-07-08T09:08:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,454,286 | 2,607 | ztlpn | Currently there is no command to alter table settings. So yes, detach x / fix x.sql / attach x is the correct way to do it. For replicated tables you can change settings independently for each replica (except the `index_granularity` setting which must be equal for all replicas). | 2018-07-09T11:56:38 | 2018-07-09T11:56:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,251,710 | 1,546 | alexey-milovidov | PS. You can also test with more simple test case:
- create both replicas on the same server (with different table names);
- instead of node shutdown and startup, you can use DETACH TABLE and ATTACH TABLE (the effect should be the same); | 2018-07-08T00:18:18 | 2018-07-08T00:18:18 | {} | MEMBER |
ClickHouse | ClickHouse | 403,252,986 | 2,281 | alexey-milovidov | Now I'll create 100 000 MergeTree tables, all in parallel in 16 threads:
```
seq 10001 100000 | sed -r -e 's/^.+$/CREATE TABLE multiple_mergetrees.table\0 (x UInt8) ENGINE = MergeTree ORDER BY x;/' | clickhouse-benchmark -c16
``` | 2018-07-08T00:48:43 | 2018-07-08T00:48:43 | {} | MEMBER |
ClickHouse | ClickHouse | 403,258,839 | 2,281 | alexey-milovidov | After this commit 3445c78, CPU usage for 100 000 tables drops to less than 10% of one CPU core. | 2018-07-08T03:16:57 | 2018-07-08T03:16:57 | {} | MEMBER |
ClickHouse | ClickHouse | 403,271,140 | 2,605 | blinkov | It makes sense to add those some day, but at the moment decimal/numeric seems to be in much higher demand. | 2018-07-08T08:17:04 | 2018-07-08T08:17:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,276,173 | 2,608 | arduanov | Собралось с `USE_INTERNAL_LLVM_LIBRARY=1` и `LLVM Compiler-7.0.0bundled`, но размер бинарника clickhouse очень большой - 664 мегабайта.
| 2018-07-08T09:54:15 | 2018-07-08T09:54:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,571,296 | 2,266 | alexey-milovidov | Now the build work smoothly (if you build directly on ARM64 machine).
@pavel-odintsov
Do you have any results on your servers? | 2018-07-09T18:13:40 | 2018-07-09T18:15:21 | {} | MEMBER |
ClickHouse | ClickHouse | 404,078,667 | 2,156 | jpiper | This is still happening for me on 1.1.54388 :( | 2018-07-11T07:53:52 | 2018-07-11T07:53:52 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 404,196,080 | 1,656 | alex-zaitsev | Supporting TRUNCATE for Set tables would be useful as well. | 2018-07-11T14:45:48 | 2018-07-11T14:45:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,072,324 | 2,281 | alexey-milovidov | > Should I add the interval in a config setting ? or I can hardcode between 8 and 12?
You can just put constant in code, similar to
```
static const auto MERGE_SELECTING_SLEEP_MS = 5 * 1000;
```
> For my personal knowledge can you please elaborate a bit why "It's important to select merges even when the... | 2018-07-06T15:47:11 | 2018-07-06T15:47:11 | {} | MEMBER |
ClickHouse | ClickHouse | 403,627,788 | 2,617 | simPod | @alexey-milovidov oh I see. I tried to break the test and the build passed so I though it's not run.
Probably the issue is just error code 0 when there is an error | 2018-07-09T21:34:26 | 2018-07-09T21:34:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,619,435 | 1,939 | fr05t1k | @PangKuo @harlinb's suggestion has worked for me. | 2018-07-12T19:10:42 | 2018-07-12T19:10:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,037,545 | 2,111 | silviucpp | But creation of the array shouldn't somehow find the best interval that can fit all the values ?
Don't you think this will be the best approach ?
Silviu | 2018-07-06T13:40:44 | 2018-07-06T13:40:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,251,868 | 1,546 | alexey-milovidov | > node1 has 4 items in system.replication_queue
The most important - what is the reason for these items failed to execute.
You should look either to:
- exception;
- postpone reason;
- `thead apply all backtrace` in gdb, if the execution of queue items was not even attempted.
Also it's important if you have ha... | 2018-07-08T00:22:39 | 2018-07-08T00:22:39 | {} | MEMBER |
ClickHouse | ClickHouse | 403,252,884 | 2,281 | alexey-milovidov | Now lets insert something into all of these tables:
```
seq 1 10000 | sed -r -e 's/^.+$/INSERT INTO multiple_mergetrees.table\0 VALUES (0)/' | clickhouse-client -n
```
It takes only 5 seconds to execute 10 000 inserts.
After insertion, there is no increase in CPU usage. And still no spikes. | 2018-07-08T00:46:10 | 2018-07-08T00:46:10 | {} | MEMBER |
ClickHouse | ClickHouse | 403,258,819 | 2,281 | alexey-milovidov | I cannot reproduce the issue that you describe.
Maybe it exists if you run debug build. Or if you run under Mac OS. | 2018-07-08T03:16:24 | 2018-07-08T03:16:24 | {} | MEMBER |
ClickHouse | ClickHouse | 403,436,783 | 2,608 | proller | когда не собралось - как точно был запущен cmake ?
и если можно - что вывел cmake в консоль? | 2018-07-09T10:41:07 | 2018-07-09T10:41:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,733,675 | 2,611 | amosbird | Weird, I can't reproduce it. What's the table structure. Could you post the full reproducible steps? | 2018-07-10T07:53:18 | 2018-07-10T07:53:18 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 403,797,977 | 2,611 | simPod | Ok got it. Happens when column is in primary key
```sql
CREATE TABLE IF NOT EXISTS test
(
created_at DateTime,
id UUID
)
ENGINE = MergeTree
PARTITION BY toDate(created_at)
ORDER BY (created_at, id);
INSERT INTO test (created_at, id) VALUES (now(), '00000000-0000-03f8-9cb8-cb1b82fb3900');
SELE... | 2018-07-10T12:02:59 | 2018-07-10T12:02:59 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,833,738 | 2,623 | filimonov | There was few issues with IN, and may by that case was already fixed in last versions.
But I can't reproduce even on same version.
With `WHERE ids IN ( SELECT 116 )` i get an exception about incompatible types, so i need to do typecast manually.
```
:) create table test Engine=MergeTree ORDER BY number as S... | 2018-07-10T14:02:52 | 2018-07-10T14:03:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,168,830 | 2,560 | silviucpp | I agree, anyway the bug can be replicated only with remote_function, without works as expected. | 2018-07-11T13:28:24 | 2018-07-11T13:28:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,541,978 | 2,633 | alexey-milovidov | It can be implemented by adding various format options.
Possible options can be `null_as_default`.
We also have demand for similar options in CSV, like `empty_as_default`.
This is also loosely related to #1130 | 2018-07-12T14:57:59 | 2018-07-12T14:57:59 | {
"+1": 2
} | MEMBER |
ClickHouse | ClickHouse | 403,038,785 | 2,478 | silviucpp | You can close the bug as time documentation will be updated to reflect this.
Also when I checked docs few weeks ago there was no mention for WITH NAME which is a nice feature.
Thanks
Silviu | 2018-07-06T13:45:17 | 2018-07-06T13:45:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,094,382 | 2,281 | silviucpp | @filimonov Yes exactly. This is my point.
So basically after each new write the merge event is scheduled (as it is now) and that event schedules itself until returns that there is no more work to be run on that table (My assumption here is that only a new write can generate new potential work, so no more need for r... | 2018-07-06T17:12:24 | 2018-07-06T17:13:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,134,974 | 2,111 | alexey-milovidov | > Don't you think this will be the best approach?
Not quite, because it will break some other expectations.
Array, internally, can be created in two ways:
1. Simple way, where it's an array of literals: `[1, 2, 3]`.
2. Complex way, where it's an array of more complex expressions: `[1 + 2, 3]`.
In the first c... | 2018-07-06T20:12:12 | 2018-07-06T20:13:32 | {} | MEMBER |
ClickHouse | ClickHouse | 403,509,185 | 2,614 | alexey-milovidov | Спасибо, всё отлично! | 2018-07-09T14:59:28 | 2018-07-09T14:59:28 | {} | MEMBER |
ClickHouse | ClickHouse | 404,077,647 | 2,322 | filimonov | Fixed in https://github.com/yandex/ClickHouse/releases/tag/v1.1.54388-stable | 2018-07-11T07:49:45 | 2018-07-11T07:49:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,120,447 | 2,266 | proller | about sse errors - can you show full cmake output?
your errors is just failed tests:
```-- Performing Test HAVE_SSE41
-- Performing Test HAVE_SSE41 - Failed
-- Performing Test HAVE_SSE42
-- Performing Test HAVE_SSE42 - Failed
-- Performing Test HAVE_POPCNT
-- Performing Test HAVE_POPCNT - Failed``` | 2018-07-11T10:22:14 | 2018-07-11T10:22:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,240,083 | 2,407 | alexey-milovidov | I don't understand, what's wrong with `keep_alive_timeout` - it's exactly for tuning timeout of HTTP/1.1 Connection: Keep-Alive.
Ok, I will close. | 2018-07-11T16:57:07 | 2018-07-11T16:57:27 | {} | MEMBER |
ClickHouse | ClickHouse | 403,078,625 | 2,281 | filimonov | As i understand @silviucpp propose to stop triggering background merges if previous background merge run found that there are no parts to merge (everything that could be merged was already merged)
I.e. there was 10 parts
1) first background merge merged 8 of them, and save flag "some work was done, more merges can ... | 2018-07-06T16:10:54 | 2018-07-06T16:16:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,569,746 | 2,299 | alexey-milovidov | You can run
```
clickhouse-client --host=127.0.0.1
```
instead of
```
clickhouse-client
``` | 2018-07-09T18:08:16 | 2018-07-09T18:08:16 | {
"+1": 13
} | MEMBER |
ClickHouse | ClickHouse | 403,654,853 | 2,617 | alexey-milovidov | We have two Travis jobs for each commit.
This is for your pull request:
https://travis-ci.org/yandex/ClickHouse/builds/401968092?utm_source=github_status&utm_medium=notification
Second job failed as expected:
https://travis-ci.org/yandex/ClickHouse/jobs/401968094
Although the first job passed. As I remember,... | 2018-07-09T23:50:42 | 2018-07-09T23:51:02 | {} | MEMBER |
ClickHouse | ClickHouse | 404,221,327 | 1,959 | alexey-milovidov | > ```\x01\e[1;31;42m :)\e[0m\x02```
I think it should be
```\x01\e[1;31;42m\x02 :)\x01\e[0m\x02``` | 2018-07-11T15:55:55 | 2018-07-11T15:55:55 | {} | MEMBER |
ClickHouse | ClickHouse | 403,114,929 | 2,281 | alexey-milovidov | Selecting parts to merge is dependent on age of parts.
See `SimpleMergeSelector`.
For example, some parts may be not selected for merge right now, but they will be selected after few minutes, even if the set of parts was not changed. | 2018-07-06T18:37:34 | 2018-07-06T18:37:34 | {} | MEMBER |
ClickHouse | ClickHouse | 403,121,908 | 2,115 | silviucpp | Ok, so problem is in another place. | 2018-07-06T19:08:44 | 2018-07-06T19:08:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,252,691 | 2,281 | alexey-milovidov | This script creates 10 000 MergeTree tables sequentially:
```
clickhouse-client --query="CREATE DATABASE multiple_mergetrees"
seq 1 10000 | sed -r -e 's/^.+$/CREATE TABLE multiple_mergetrees.table\0 (x UInt8) ENGINE = MergeTree ORDER BY x;/' | clickhouse-client -n
```
(but doesn't write anything into these tables)... | 2018-07-08T00:41:35 | 2018-07-08T00:41:35 | {} | MEMBER |
ClickHouse | ClickHouse | 403,792,238 | 1,567 | valyala | FYI, [LocustDB](https://github.com/cswinter/LocustDB) implemented dictionary encoding for string column blocks with low cardinality. This gave 50x speedup comparing to clickhouse on the following query on [New York taxi rides dataset](http://www.nyc.gov/html/tlc/html/about/trip_record_data.shtml):
```sql
SELECT
... | 2018-07-10T11:37:32 | 2018-07-10T11:37:32 | {
"+1": 4
} | NONE |
ClickHouse | ClickHouse | 403,794,550 | 2,611 | simPod | You're right. Just tried to create minimal example and it works. Will try to dig the cause. | 2018-07-10T11:47:49 | 2018-07-10T11:47:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,799,781 | 1,567 | alexey-milovidov | Task description: https://gist.githubusercontent.com/alexey-milovidov/ebdc6a0b8731fdba9438bda3ec6e8ca4/raw/8321be55ec0e63ed5aaec28b702aa52c280eb87c/column_with_dictionary_task.txt | 2018-07-10T12:10:33 | 2018-07-10T12:10:33 | {} | MEMBER |
ClickHouse | ClickHouse | 404,168,259 | 2,407 | silviucpp | Feel free to close this one as time is not a problem. I've tested only on mac os. | 2018-07-11T13:26:49 | 2018-07-11T13:26:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,169,687 | 1,656 | silviucpp | You think this is a good feature ? if not we can close it. I personally don't use it any longer. not sure if others needs it. | 2018-07-11T13:30:49 | 2018-07-11T13:30:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,268,873 | 1,322 | alexey-milovidov | Yes, it isn't fixed. | 2018-07-11T18:35:26 | 2018-07-11T18:35:26 | {} | MEMBER |
ClickHouse | ClickHouse | 403,097,997 | 2,115 | silviucpp | @alexey-milovidov you consider this a valid bug ? | 2018-07-06T17:26:52 | 2018-07-06T17:26:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,623,554 | 2,617 | simPod | Also, it passes when there's a build error
```
1 error generated.
[796/1396] Building CXX object contrib/poco/Data/CMakeFiles/Data.dir/src/StatementImpl.cpp.o
ninja: build stopped: subcommand failed.
+date
Mon Jul 9 21:06:43 UTC 2018
The command "utils/travis/normal.sh" exited with 0.
``` | 2018-07-09T21:18:00 | 2018-07-09T21:18:00 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,798,991 | 1,567 | alexey-milovidov | @KochetovNicolai is doing this task in
https://github.com/yandex/ClickHouse/tree/data-type-with-dictionary
branch. | 2018-07-10T12:07:21 | 2018-07-10T12:07:21 | {} | MEMBER |
ClickHouse | ClickHouse | 404,195,129 | 1,959 | filimonov | readline doesn't know that prompt string contains ANSII escape sequences, so it count unprintable chars while calculating line width. See https://wiki.hackzine.org/development/misc/readline-color-prompt.html
So workaround is adding special marker `\x01` at the begining and `\x02` at the end. I've tested that local... | 2018-07-11T14:43:08 | 2018-07-11T14:56:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,848,609 | 448 | blinkov | Still reproducible with recent clickhouse-server. | 2018-07-13T14:24:26 | 2018-07-13T14:24:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,068,776 | 2,595 | filimonov | > I think it's a bug when creating a column in tables of the Replicated type
Please give some details to help debug it:
1) version of Clickhouse used.
2) exact failing query. The query you've provided is not correct. It can't be parsed because of bad backspaces sequence:
```
:) ALTER TABLE Table_Rep ADD COLUMN ... | 2018-07-06T15:33:39 | 2018-07-06T15:33:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,136,078 | 2,560 | alexey-milovidov | Yes, but we need to dig deeper, as it may indicate wrong logic in code. | 2018-07-06T20:16:57 | 2018-07-06T20:16:57 | {} | MEMBER |
ClickHouse | ClickHouse | 403,253,276 | 2,281 | alexey-milovidov | Now clickhouse-server constantly eats about 10..50% of single CPU core and no spikes exist. | 2018-07-08T00:56:26 | 2018-07-08T00:56:26 | {} | MEMBER |
ClickHouse | ClickHouse | 403,653,796 | 2,618 | alexey-milovidov | That looks mostly correct.
Only few issues remain:
1. I cannot find implementation of `FieldVisitorToString` for `UInt128`. That will likely require some effort.
2. Comparison (less/greater) operators for UInt128 may work even with your implementation (that returns false when comparing with other integer type) but... | 2018-07-09T23:44:10 | 2018-07-09T23:44:10 | {} | MEMBER |
ClickHouse | ClickHouse | 404,189,853 | 1,656 | alexey-milovidov | That's good feature. Reading from backup files (with DISTINCT on top) seems fairly straightforward. | 2018-07-11T14:28:37 | 2018-07-11T14:28:37 | {} | MEMBER |
ClickHouse | ClickHouse | 404,198,578 | 1,656 | alexey-milovidov | TRUNCATE is already implemented for Set and Join. | 2018-07-11T14:52:27 | 2018-07-11T14:52:27 | {} | MEMBER |
ClickHouse | ClickHouse | 404,208,919 | 2,612 | Joozty | @filimonov I already solved it with `ARRAY JOIN` clause. Instead of using JSON dict I decided to use array like this: `[[key, value], [key, value] ...]`
It works well. The only problem I have now is with MATERIALIZED VIEW. (a similar question was posted on [Stackoverflow](https://stackoverflow.com/questions/51233488... | 2018-07-11T15:21:03 | 2018-07-11T15:21:03 | {} | NONE |
ClickHouse | ClickHouse | 404,220,697 | 1,959 | alexey-milovidov | Ok. That means, we only have to update example. | 2018-07-11T15:54:09 | 2018-07-11T15:54:09 | {} | MEMBER |
ClickHouse | ClickHouse | 404,229,580 | 2,612 | Joozty | I tried to use arrays to avoid `JOIN` but I wasn't successful because there is no possibility to use `UNION ALL` in materialized views. And I needed it.
To show you what I am trying to achieve.
The table contains `ID, one_array, second_array`. These arrays are in following format `[[int, int], [int, int]...]`.... | 2018-07-11T16:21:51 | 2018-07-11T16:21:51 | {} | NONE |
ClickHouse | ClickHouse | 404,429,021 | 2,546 | dvska | and for **clickhouse-client** please | 2018-07-12T08:14:58 | 2018-07-12T08:14:58 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,852,692 | 2,021 | blinkov | Let's keep just #253 for this. | 2018-07-13T14:38:06 | 2018-07-13T14:38:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,065,631 | 2,334 | filimonov | > Multi-statements are not allowed
Remove semicolon from the query end. ;) | 2018-07-06T15:22:04 | 2018-07-06T15:22:04 | {
"heart": 1,
"hooray": 4
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,126,202 | 1,959 | alexey-milovidov | Colorful prompt doesn't work properly.
Example:
1. Run clickhouse-client with whatever colorized prompt.
2. Type 'aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa'.
3. Press Home.
| 2018-07-06T19:29:26 | 2018-07-06T19:29:26 | {} | MEMBER |
ClickHouse | ClickHouse | 403,129,909 | 2,115 | alexey-milovidov | Fix is in master. | 2018-07-06T19:47:38 | 2018-07-06T19:47:38 | {} | MEMBER |
ClickHouse | ClickHouse | 403,184,226 | 2,579 | alexey-milovidov | > So as I understand, we can not use the same table. We need to create a new table in the new cluster, then copy data to this table and then drop old table (if we need it) ?
Yes.
Also you can create a new table on the same cluster.
But resharding within the same table on the same cluster is not possible.
| 2018-07-07T03:05:57 | 2018-07-07T03:05:57 | {} | MEMBER |
ClickHouse | ClickHouse | 403,236,598 | 2,563 | alexey-milovidov | Исправление в master. | 2018-07-07T19:02:38 | 2018-07-07T19:02:38 | {} | MEMBER |
ClickHouse | ClickHouse | 403,236,672 | 2,587 | alexey-milovidov | Не хватает структуры таблицы и данных для воспроизведения. | 2018-07-07T19:04:13 | 2018-07-07T19:04:13 | {} | MEMBER |
ClickHouse | ClickHouse | 403,550,100 | 2,606 | shawel | Thanks @KochetovNicolai for the response.
> If you know the exactly number of columns you need to store, try to use tuple
The array length is very arbitrary (~100 average length) so tuple might not work.
> You need to settle a condition on full array in order to get profit from pk
In my use case i have t... | 2018-07-09T17:06:42 | 2018-07-09T17:06:42 | {} | NONE |
ClickHouse | ClickHouse | 403,997,004 | 2,266 | pavel-odintsov | Hello!
Sorry, I do not have any updates about it. Load on this machine is very small :( | 2018-07-10T23:26:34 | 2018-07-10T23:26:34 | {} | NONE |
ClickHouse | ClickHouse | 404,216,335 | 2,612 | Joozty | @filimonov Is it possible to refresh a materialized view manually at least? | 2018-07-11T15:41:40 | 2018-07-11T15:41:40 | {} | NONE |
ClickHouse | ClickHouse | 404,261,929 | 1,322 | silviucpp | @alexey-milovidov Also this one I'm not sure was not fixed in the meantime. I didn't checked on the master but seems nobody reviewed it to mark it as bug or close it. | 2018-07-11T18:11:29 | 2018-07-11T18:11:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,129,885 | 2,560 | silviucpp | While investigating this issue I found that problem is in InterpreterSelectQuery line 129 where getQueryContext() point to an already released context.
To avoid such problems maybe it's a better idea to use them as shared_ptr ? | 2018-07-06T19:47:29 | 2018-07-06T19:47:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,246,868 | 2,605 | alex-zaitsev | PostgreSQL network types: https://www.postgresql.org/docs/9.1/static/datatype-net-types.html | 2018-07-07T22:29:33 | 2018-07-07T22:29:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,213,745 | 2,612 | filimonov | > @filimonov I already solved it with ARRAY JOIN clause. Instead of using JSON dict I decided to use array like this: [[key, value], [key, value] ...]
Also array of tuples also could be a nice solution.
Like
```
SELECT
dict.1 AS key,
dict.2 AS value
FROM system.one
ARRAY JOIN [('key', 'value'), ('... | 2018-07-11T15:34:18 | 2018-07-11T15:34:41 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,116,991 | 2,281 | silviucpp | OK this is the detail that makes the difference :) Then only improvement we can make is to apply the exponential backoff | 2018-07-06T18:46:49 | 2018-07-06T18:46:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 403,264,302 | 2,281 | alexey-milovidov | > I see periodic stalls for 10s of seconds.
Stalls are in Linux kernel:
```
sudo strace -f -T -p $(pidof clickhouse-server) 2>&1 | grep -v futex | grep -P '<[1-9]'
[pid 663052] <... open resumed> ) = 8 <59.698087>
[pid 241203] <... unlink resumed> ) = 0 <60.287390>
[pid 241202] <... unlink resumed... | 2018-07-08T05:46:05 | 2018-07-08T05:48:17 | {} | MEMBER |
ClickHouse | ClickHouse | 403,425,504 | 2,606 | KochetovNicolai | Arrays can be used as a primary key. However, arrays are stored continuously as a pair of columns with values and offsets. So, primary key can't be used for single array element. You need to settle a condition on full array in order to get profit from pk.
If you know the exactly number of columns you need to store, ... | 2018-07-09T09:55:47 | 2018-07-09T09:55:47 | {} | MEMBER |
ClickHouse | ClickHouse | 403,626,196 | 2,617 | alexey-milovidov | It certainly not passes when there is a build error or test fail.
@proller ? | 2018-07-09T21:28:12 | 2018-07-09T21:28:12 | {} | MEMBER |
ClickHouse | ClickHouse | 404,175,852 | 2,629 | filimonov | Confirmed.
test case:
```
echo '<yandex><user>user_from_config</user></yandex>' > test_user_from_config.xml
clickhouse-client -c test_user_from_config.xml --query='SELECT 1' | grep 'Unknown user'
```
For host - that work properly.
```
echo '<yandex><host>host_from_config</host></yandex>' > test_host_from_c... | 2018-07-11T13:48:07 | 2018-07-11T13:48:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 404,206,140 | 2,612 | filimonov | What do you mean by DICTIONARY? JSON field? | 2018-07-11T15:13:27 | 2018-07-11T15:13:27 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.