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
488,448,428
5,095
alexey-milovidov
IPv6CIDRtoIPv6Range -> IPv6CIDRToRange
2019-05-01T22:04:06
2019-05-01T22:04:06
{}
MEMBER
ClickHouse
ClickHouse
489,610,797
5,203
ghost
@den-crane Thank you, it works even better!
2019-05-06T12:58:16
2019-05-06T12:58:16
{}
NONE
ClickHouse
ClickHouse
488,872,605
5,176
abraithwaite
> double start/and stop clickhouse does not sound like a brilliant idea in such a production usecase. Totally agree. But why even bother sourcing non-executable scripts if not to set the environment for clickhouse-server before it starts? That's what I don't understand. https://hub.docker.com/r/yandex/clickh...
2019-05-02T23:54:38
2019-05-02T23:54:38
{}
NONE
ClickHouse
ClickHouse
489,462,373
5,181
danlark1
https://github.com/yandex/ClickHouse/pull/5198
2019-05-05T20:52:27
2019-05-05T20:52:27
{ "+1": 1 }
CONTRIBUTOR
ClickHouse
ClickHouse
489,621,342
5,201
MrBuddyCasino
I just tested this, and executed `select countMerge(row_count_state) from calls_fast_row_count` after running the scripts. It returned the correct result. I made an updated version of the zip, get it at https://drive.google.com/file/d/1A53kAmMeOHzAzYKaLo_yRLhxzjdpsZ0o/view?usp=sharing.
2019-05-06T13:31:47
2019-05-06T13:31:47
{}
NONE
ClickHouse
ClickHouse
489,643,555
5,205
bookin
@den-crane, I understood, thank for help
2019-05-06T14:34:38
2019-05-06T14:34:38
{}
NONE
ClickHouse
ClickHouse
488,705,631
5,166
den-crane
> What's the risk here? Aside from potential slowness Problems with ZK transactions. To start a mutation CH performs a ZK transaction which includes a list with all partitions names and ZK transaction has limited length (in bytes). Basically is impossible to execute alter delete/update if the table has more than 1...
2019-05-02T14:54:44
2019-05-02T14:56:24
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,819,782
5,176
abraithwaite
Our environment variables for macros aren't being sourced before clickhouse quits. We can't pass env through docker because we want the same configuration for each server in kubernetes (we're using a statefulset). What this resulted in was us extending this image and copying in the few environment files we need to ...
2019-05-02T20:26:59
2019-05-02T20:30:10
{}
NONE
ClickHouse
ClickHouse
489,472,478
4,860
alexey-milovidov
@den-crane > I don't really understand what is the difference if you store a seconds' fraction in another column? > Just convince for in your backend? 1. Better compression of integer part. 2. A room for optimization if we will be able to read just integer part.
2019-05-05T23:22:13
2019-05-05T23:22:13
{}
MEMBER
ClickHouse
ClickHouse
489,805,981
2,103
filimonov
Can you recheck that on the latest versions? (I was not able to reproduce that)
2019-05-06T22:27:37
2019-05-06T22:27:37
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,656,597
5,095
alexey-milovidov
I will fix all the remaining issues.
2019-05-02T12:40:09
2019-05-02T12:40:09
{}
MEMBER
ClickHouse
ClickHouse
488,819,079
5,177
kmatt
Using DEFAULT instead of MATERIALIZED results in same.
2019-05-02T20:24:39
2019-05-02T20:24:39
{}
NONE
ClickHouse
ClickHouse
489,183,110
5,185
den-crane
Distributed table sends the same query to all shards (because it does not know that a cluster never changed (number of nodes or sharding key) and data inserted accordingly with a sharding rule). So in case of [ **prefer_localhost_replica** = 0, **max_parallel_replicas** = 2 ] Distributed table requests 2 replicas and ...
2019-05-03T17:51:02
2019-05-03T17:51:02
{ "+1": 1 }
CONTRIBUTOR
ClickHouse
ClickHouse
489,609,143
5,203
den-crane
Just in case. Are you aware of an optimization `PRIMARY_KEY_COLUMN IN ()` including `PRIMARY_KEY_COLUMN IN (SUBQUERY)` SELECT count() FROM table WHERE CounterId in (123, 345, ....) For long lists with enumerations check https://clickhouse.yandex/docs/en/operations/table_engines/external_data/
2019-05-06T12:52:42
2019-05-06T12:54:57
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,828,858
5,199
filimonov
ClickHouse supports 3 consistent hash algorithms: ``` SELECT * FROM system.functions WHERE name LIKE '%Consi%' ┌─name─────────────────┬─is_aggregate─┬─case_insensitive─┬─alias_to─┐ │ sumburConsistentHash │ 0 │ 0 │ │ │ jumpConsistentHash │ 0 │ ...
2019-05-06T23:55:38
2019-05-06T23:55:57
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,530,706
5,160
notbdu
That was it, thanks @den-crane and @blinkov! I added this to the `/etc/clickhouse-server/users.xml` and it is working now :).
2019-05-02T01:38:09
2019-05-02T01:38:09
{}
NONE
ClickHouse
ClickHouse
488,965,767
4,933
helloIAmPau
Any update on this? The `arrayGroupBy` proposal by @andrea-s sounds very interesting to me and could be a perfect solution for a couple of use cases I'm currently working on.
2019-05-03T06:39:44
2019-05-03T06:39:44
{}
NONE
ClickHouse
ClickHouse
489,198,501
5,186
den-crane
Be aware that `FROM FINAL` is not usable with the real FACT table (which needs max_parallel_replicas ). In case of billions rows FINAL is super slow.
2019-05-03T18:39:50
2019-05-03T18:39:50
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,457,225
4,860
aadant
Datetime(6), timestamp(6), Datetime(3), timestamp(3) are supported by databases like MySQL, Postgres, Redshift, MariaDB column store, SAP IQ. For nanoseconds, people use an unsigned bigint (10 for epoch and 9 for nanoseconds).
2019-05-05T19:43:28
2019-05-05T19:43:28
{}
NONE
ClickHouse
ClickHouse
489,462,448
5,197
den-crane
``` | => cat c.sh #!/bin/bash clickhouse-client -q 'drop table if exists jt' sleep 1 clickhouse-client -q 'create table jt (A Int64, S0 String, S1 String, S2 String, S3 String) Engine=MergeTree order by A' sleep 1 (while [ 1 ]; do clickhouse-client --use_uncompressed_cache=1 -q 'select count() from jt left j...
2019-05-05T20:53:37
2019-05-05T20:53:37
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,462,133
5,187
filimonov
> In straightforward implementation (a number of nanoseconds since the epoch), Int64 data type will give us about 292 years around 1970: BTW: anyway DateLUT should be adjusted to support that, right? Bacause it will start getting Int64 instead of UInt32.
2019-05-05T20:48:56
2019-05-05T20:59:10
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,463,493
5,197
danlark1
Is it reproducible with later clickhouse builds like v19.7.1.403 ?
2019-05-05T21:08:16
2019-05-05T21:09:12
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,468,270
5,042
alexey-milovidov
At least 164 tests have failed.
2019-05-05T22:15:11
2019-05-05T22:15:11
{}
MEMBER
ClickHouse
ClickHouse
489,794,453
2,389
filimonov
The issue is still here. Simpler test case: ```sql CREATE TABLE session ( day Date, uid String, dummy String DEFAULT '' ) ENGINE = MergeTree(day, (day, uid), 32); CREATE TABLE queue ( day Date, uid String ) ENGINE = MergeTree(day, (day, uid), 32); CREATE MATERIALIZED VIEW IF NOT EXI...
2019-05-06T21:43:31
2019-05-06T21:43:50
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,860,563
5,174
filimonov
On 19.4 i get the following exception: ``` Received exception from server (version 19.4.2): Code: 182. DB::Exception: Received from localhost:9000, 127.0.0.1. DB::Exception: Storage Buffer doesn't support PREWHERE.. ``` On 19.5 weird exception as in issue ``` Received exception from server (version 19.5.2): ...
2019-05-07T00:32:37
2019-05-07T00:32:37
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,851,332
5,172
alexey-milovidov
The test didn't find any issues.
2019-05-02T22:14:01
2019-05-02T22:14:01
{}
MEMBER
ClickHouse
ClickHouse
489,111,390
5,182
nvartolomei
Can’t plus this enough. I would propose for open census as a two in one solution for both metrics and tracing, although I don’t have experience with it’s cpp client.
2019-05-03T14:19:04
2019-05-03T14:19:04
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,301,738
5,182
alexvaut
Would love to this that available in CH ! On .NET, the open tracing client is very well done and easy to integrate, of course it supports sampling which is required in many contexts where clickhouse is used. Tracing is one of the feature that was coming for free in monolith app time. Now, in the reign of microservi...
2019-05-04T07:04:01
2019-05-04T07:04:01
{}
NONE
ClickHouse
ClickHouse
489,438,570
5,116
alexey-milovidov
> Thanks for the feedback, does it's possible to have more information about how you were able to produce the bugs ? Tried to do a simple test and wasn't able to reproduce it :/ I did not done it manually, just clicked to "Details" link near the checks, then clicked to relevant logs.
2019-05-05T15:52:10
2019-05-05T15:52:19
{}
MEMBER
ClickHouse
ClickHouse
489,464,448
5,197
danlark1
And if you can, provide `sudo strace -fe mmap -p $CHPID 2>&1 | grep "mmap(0"` of the server if you can
2019-05-05T21:21:17
2019-05-05T21:21:17
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,546,233
5,201
MrBuddyCasino
These issues might be related: https://github.com/yandex/ClickHouse/issues/2878 https://github.com/yandex/ClickHouse/issues/956
2019-05-06T08:31:01
2019-05-06T08:31:01
{}
NONE
ClickHouse
ClickHouse
489,611,034
5,201
den-crane
Materialized view is INSERT_TRIGGER. Engine Merge does not support INSERTs so MV over Merge is nonsense. MV subscribes on INSERT for a table. When a table gets INSERT it pushes INSERTED_BUFFER to the MV. MV never reads the table directly so the table could be a fake (engine=Null) and does not store data.
2019-05-06T12:59:03
2019-05-06T13:01:29
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,629,741
5,201
den-crane
>It returned the correct result. haha, just make your inserts in the end, after MV creation.
2019-05-06T13:56:31
2019-05-06T13:56:31
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,775,452
5,201
filimonov
The bug here is that is allowed you to create the Materialized View which has no practical sense. If you need to collect data from several tables to one materialized view you do something like: ```sql create table source1(...) Engine=...; create table source2(...) Engine=...; create table target (...) Engine=....
2019-05-06T21:02:12
2019-05-06T21:02:12
{ "+1": 1 }
CONTRIBUTOR
ClickHouse
ClickHouse
489,801,736
3,529
filimonov
Zookeeper behaves bad. Maybe it's installed on same machines as ClickHouse? (it's bad practice) > Since MV works as triggers (processing asynchronously after initial query) I think there will be no any difference. MV is executed synchronously, it's Distributed to ReplicatedMergeTree insert work asynchronously. T...
2019-05-06T22:10:11
2019-05-06T22:10:11
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,821,384
3,925
vzakaznikov
I finally got permission to participate in ClickHouse open-source project on my free time and so if there is interest from your side we could try to revive this PR. I will try to provide the supporting discussion related to LIVE VIEWs and LIVE CHANNELs. Hopefully, it can be useful and will revive thе interest in thi...
2019-05-06T23:17:49
2019-05-06T23:17:49
{ "+1": 3 }
CONTRIBUTOR
ClickHouse
ClickHouse
488,200,042
751
abraithwaite
Following up here, in addition to the data being inconsistent in the result, it's also completely nonsensical and should throw an error. Take below for example. We're filtering on hour, but incorrectly were grouping by more than what's in our sorting key: `(sourceID, hour)`. The first query correctly returns no resu...
2019-05-01T03:30:39
2019-05-01T22:33:22
{ "+1": 2 }
NONE
ClickHouse
ClickHouse
488,653,174
5,166
andrea-s
What's the risk here? Aside from potential slowness
2019-05-02T12:28:21
2019-05-02T12:28:21
{}
NONE
ClickHouse
ClickHouse
488,969,372
4,554
yellow1912
@4ertus2 I think I run into the same issue here? https://github.com/yandex/ClickHouse/issues/5180 I'm on 19.5.3, I wonder if I need to update?
2019-05-03T06:49:52
2019-05-03T06:49:52
{}
NONE
ClickHouse
ClickHouse
489,564,471
1,370
TheRoSS
The bug is still here for distributed tables (19.5.3.8) ``` > CREATE TABLE gb_refs.currencyRateShard (`date` Date, `base` FixedString(3), `code` FixedString(3), `src` Int8, `rate` Decimal(18, 6)) ENGINE = ReplicatedReplacingMergeTree('/clickhouse/tables/{shard}/gb_refs/currencyRateShard', '{replica}') PARTITION BY ...
2019-05-06T09:43:45
2019-05-06T09:43:45
{}
NONE
ClickHouse
ClickHouse
489,608,418
5,201
MrBuddyCasino
Turns out this is probably a regex issue. The merge table definition should have used `^calls$|^longterm_calls$|^calls_v2$`, but was `'^calls|longterm_calls|calls_v2$'`
2019-05-06T12:50:09
2019-05-06T12:50:24
{}
NONE
ClickHouse
ClickHouse
489,614,360
5,030
YiuRULE
@proller Update tests as well, does we need to apply other change ? I think I cover everything now
2019-05-06T13:10:13
2019-05-06T13:10:13
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,498,503
5,160
den-crane
https://clickhouse.yandex/docs/en/operations/settings/settings/#connect-timeout-with-failover-ms connect_timeout_with_failover_ms Default value: 50. cat /etc/clickhouse-server/conf.d/user_substitutes.xml ``` <?xml version="1.0"?> <yandex> <profiles> <default> <connect_timeout_with_fail...
2019-05-01T23:42:00
2019-05-01T23:42:00
{ "+1": 4 }
CONTRIBUTOR
ClickHouse
ClickHouse
488,627,087
5,168
blinkov
cc @zhang2014 @sundy-li
2019-05-02T10:41:27
2019-05-02T10:41:27
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,668,286
247
alexey-milovidov
> Yes, I could store this data in two fields but it does not look as great solution. Also I want to have ability to extract / handle this data with CH bundled functions without external additional processing. The functions have been added in #5095 About storage of IP address and mask length in separate columns - ...
2019-05-02T13:15:34
2019-05-02T13:15:34
{}
MEMBER
ClickHouse
ClickHouse
488,811,948
5,176
filimonov
docker-entrypoint-initdb.d is intended to initialize the database, so it starts clickhouse, run the init scripts, stops clickhouse, and after that run it again "normally". Can I ask why do you need to run some scripts before that? Why can't you use regular way of passing env to docker?
2019-05-02T20:02:38
2019-05-02T20:03:04
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,059,200
5,180
4ertus2
Your variant of table aliasing is not supported. 'alias as e1' is not visible here cause it's in subselect. You should place aliases for subselect themselfs, not for tebles inside subselects. ``` select ... from () e1 any left join () e2 on e2.user_id = e1.user_id ... ```
2019-05-03T11:00:11
2019-05-03T11:00:11
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,344,189
5,072
4ertus2
It's cause COMMA JOIN rewriter do not understand IN clause (and many other patterns) yet. So it rewrites query to CROSS JOIN instead of INNER one.
2019-05-04T16:52:55
2019-05-04T16:52:55
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,431,540
5,190
den-crane
Test (19.7.1.402): create table jt(A Int64, S0 String, S1 String, S2 String, S3 String) Engine=MergeTree order by A; insert into jt select number, toString(arrayMap(x->x, range(100))) s, s, s, s from numbers(500000); select count() from jt left join jt using A; Elapsed: 0.309 sec. select count() from jt left ...
2019-05-05T14:28:10
2019-05-05T14:33:30
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,456,755
4,860
filimonov
> @filimonov : MySQL and Postgres natively support datetime(6) and timestamp(6) that have microsecond precision. If you want to replicate them, you need to support this data type. I think MariaDB column store just added it. https://mariadb.com/kb/en/library/columnstore-data-types/ I've wrote above: "(DateTime64 is) ...
2019-05-05T19:37:11
2019-05-05T21:43:26
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,472,466
5,197
den-crane
I could not reproduce this with ~master 19.8.1.438 & 19.8.1.443 & with testing 19.7.1.403 Though my test is not very stable , I'll try to get more reproducible results.
2019-05-05T23:22:08
2019-05-05T23:22:08
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,896,148
5,199
etah000
I see. The computation of the reminder does not influence the data distribution generated by "jumpConsistentHash". In order to follow the data distribution weight specified by shard configuration parameter, the "bucket number" should be the total weight of all the shards. Thanks.
2019-05-07T03:45:13
2019-05-07T03:45:13
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,464,244
5,190
alexey-milovidov
@zhang2014 has fixed the issue.
2019-05-05T21:18:23
2019-05-05T21:18:23
{}
MEMBER
ClickHouse
ClickHouse
489,700,705
4,282
kmatt
> Is the issue anyhow related to the Kafka table? > (Does it reproduce if you first remove the Kafka table and materialized view and only then issue an ALTER command?) > > Could you please install `clickhouse-common-dbg` package, reproduce the issue and run > `sudo gdb -batch -ex 't apply all bt' -p $(pidof clickh...
2019-05-06T17:18:41
2019-05-06T17:18:41
{}
NONE
ClickHouse
ClickHouse
489,811,877
1,993
filimonov
relates #5207 (about better exception) #4006 (some ideas of implementation something similar).
2019-05-06T22:33:55
2019-05-06T22:33:55
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,821,003
5,165
filimonov
Related: https://github.com/yandex/ClickHouse/issues/4797#issuecomment-476324427 And #2691
2019-05-06T23:15:55
2019-05-06T23:21:34
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,822,355
3,997
filimonov
@mcarbonneaux did you succeed with that?
2019-05-06T23:22:45
2019-05-06T23:22:45
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,651,998
5,139
alexey-milovidov
Totally irrelevant. Doesn't fix the issue.
2019-05-02T12:23:57
2019-05-02T12:23:57
{}
MEMBER
ClickHouse
ClickHouse
489,471,777
5,197
alexey-milovidov
It looks like it was fixed in https://github.com/yandex/ClickHouse/pull/5133
2019-05-05T23:10:16
2019-05-05T23:10:16
{}
MEMBER
ClickHouse
ClickHouse
489,472,623
5,197
alexey-milovidov
The issue that was fixed is reproduced very rarely. It will be very difficult to reproduce because it relies on OS memory management. (Don't need to waste time.)
2019-05-05T23:24:48
2019-05-05T23:25:43
{}
MEMBER
ClickHouse
ClickHouse
489,472,832
5,187
alexey-milovidov
> will give 292471 years in both directions, which is 100% enough for all possible cases. Yes. Some calendar issues will arise (for calculations on historical events), but we can just ignore them. > Another (more realistic) scenario - i have microseconds in datastream, but want to store only with 1/10 precision i...
2019-05-05T23:28:44
2019-05-05T23:28:44
{}
MEMBER
ClickHouse
ClickHouse
489,769,939
5,206
filimonov
It is not so obvious to say what behaviour is correct. If you use datatype with timezone specified you force ClickHouse to use the formatting rules for that timezone when reading/writing time from that field. That's why it was read according to tz rules you've specified in datatype. When querying you're compari...
2019-05-06T20:45:18
2019-05-06T20:45:18
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,777,170
4,250
filimonov
@nathanmarlor do you have any further questions?
2019-05-06T21:08:00
2019-05-06T21:08:00
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,825,088
2,480
filimonov
Does it still reproduce on recent versions?
2019-05-06T23:36:26
2019-05-06T23:36:26
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,373,369
5,160
alexey-milovidov
Still need to fix the issue properly on our side to avoid confusion.
2019-05-04T23:31:54
2019-05-04T23:31:54
{}
MEMBER
ClickHouse
ClickHouse
489,460,944
5,187
filimonov
> > precision should be optionated (like in Decimal datatype), because nanoseconds are needed quite rare (while micro-/milli- seconds - quite often). > > Even milliseconds forces to use at least 64bit data type instead of 32bit. Consequently, there is not much difference between milliseconds and nanoseconds resoluti...
2019-05-05T20:31:51
2019-05-05T21:46:02
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,468,509
5,095
alexey-milovidov
Missing performance test.
2019-05-01T22:42:50
2019-05-01T22:42:50
{}
MEMBER
ClickHouse
ClickHouse
489,200,432
5,181
danlark1
I will add it in the near one-two days
2019-05-03T18:45:36
2019-05-03T18:45:36
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,438,249
5,192
alexey-milovidov
Ok.
2019-05-05T15:48:22
2019-05-05T15:48:22
{}
MEMBER
ClickHouse
ClickHouse
489,438,859
5,071
alexey-milovidov
This PR looks Ok, but I'm not ready to > Return bool type in comparison/logic functions because it will introduce more complications. It means, that `Bool` type will be not the first class citizen. But it still makes sense. I need to discuss this feature with my collegues a bit more, and I think we will merge.
2019-05-05T15:55:36
2019-05-05T15:55:36
{}
MEMBER
ClickHouse
ClickHouse
489,458,681
4,860
filimonov
> Datetime(6), timestamp(6), Datetime(3), timestamp(3) are supported by databases like MySQL, Postgres, Redshift, MariaDB column store, SAP IQ. And all them have microseconds precision, not nanoseconds. > For nanoseconds, people use an unsigned bigint (10 for epoch and 9 for nanoseconds). That's exactly what...
2019-05-05T20:02:32
2019-05-05T20:06:08
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,469,636
5,197
den-crane
Seems it's a very local problem reproducible in 19.7.1.402 only.
2019-05-05T22:36:32
2019-05-05T22:36:32
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,574,718
5,116
YiuRULE
Thanks ! Now both problems are resolved. We can use now a client or a server with an outdated version without any problem, the segfault is normally resolved as well
2019-05-06T10:27:42
2019-05-06T10:27:42
{ "+1": 1 }
CONTRIBUTOR
ClickHouse
ClickHouse
489,797,982
5,125
zombiemonkey
@pikhovkin I think we've discussed the two solutions and I won't be commenting further. 1. Use CollapsingMergeTree with appropriately specced hardware will allow for real-time ingestion and reporting. Note that you have to use group by with CollapsingMergeTree to achieve real-time. You will need to either using vers...
2019-05-06T21:56:11
2019-05-06T21:56:11
{ "+1": 7 }
NONE
ClickHouse
ClickHouse
488,695,482
5,166
alexey-milovidov
- inode exhaust; - difficult to resolve this situation due to extreme degradation of filesystem performance. We assume that in 100% cases this situation indicates user misconception.
2019-05-02T14:28:43
2019-05-02T14:28:43
{}
MEMBER
ClickHouse
ClickHouse
489,204,636
5,181
danlark1
And I promise it not to be fast :)
2019-05-03T18:59:57
2019-05-03T18:59:57
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,795,901
2,174
filimonov
Cascading materialized views are nor supported for now for MergeTree family. Probably you can workaround it by creating MV to Distributed table which will insert data to MergeTree, which have another MV. Or just directly fill multiple target tables from one source table.
2019-05-06T21:48:33
2019-05-06T23:13:52
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,831,706
5,175
filimonov
It sounds like some silly error, like connecting to bad zookeeper / bad commands copy& paste. Please repeat all the steps once again. Check the commands you execute are identical (even here, in issue i see they differ - in part of PRIMARY KEY (sourceID, hour)). If if not helps - please check the content of th...
2019-05-07T00:10:24
2019-05-07T00:10:40
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,445,656
5,095
alexey-milovidov
Need to remove `.FunctionsCoding.h.swo` file.
2019-05-01T21:58:11
2019-05-01T21:58:11
{}
MEMBER
ClickHouse
ClickHouse
488,827,772
5,176
filimonov
Sorry, I'm not following. You need to run some script in a similar way to docker-entrypoint-initdb.d, but before first run of clickhouse? Right? If you had already modified the image it would be simpler also to source needed files from your modified entrypoint, as double start/and stop clickhouse does not sound ...
2019-05-02T20:51:27
2019-05-02T21:06:43
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,156,975
4,015
abraithwaite
Hitting this as well. Tried setting `insert_allow_materialized_columns` but did not help on `19.5.2 revision 54417` Edit: it looks like it's a client-specific setting? How do I set this server wide and persist it? I don't see anything in the docs.
2019-05-03T16:26:17
2019-05-03T16:43:46
{}
NONE
ClickHouse
ClickHouse
489,474,142
5,198
danlark1
Failed test is flaky :(
2019-05-05T23:49:44
2019-05-05T23:49:44
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,569,111
5,043
shangshujie365
have you try cmake3 .. -DCMAKE_INSTALL_PREFIX=/usr -DGLIBC_COMPATIBILITY=OFF -DCMAKE_BUILD_TYPE:STRING=RelWithDebInfo ? it works for me
2019-05-06T10:02:35
2019-05-06T10:02:35
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,604,944
5,045
YiuRULE
Does we need to make other changes ? would be glad to assist if needed :)
2019-05-06T12:37:21
2019-05-06T12:37:21
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,478,241
5,160
notbdu
Ah, these are test servers so I don't have any replication or clustering set up. I am just testing the `remote` function right now. I did try again w/ a recently re-imaged host and still got the same error :(. From the packet capture this is the exchange I am seeing: `host01` -> `host02` (failure): ``` host0...
2019-05-01T23:05:38
2019-05-01T23:05:38
{}
NONE
ClickHouse
ClickHouse
488,554,068
5,116
YiuRULE
Thanks for the feedback, does it's possible to have more information about how you were able to produce the bugs ? Tried to do a simple test and wasn't able to reproduce it :/
2019-05-02T05:03:28
2019-05-02T05:03:28
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,805,156
3,319
abraithwaite
This issue should be closed.
2019-05-02T19:40:39
2019-05-02T19:40:39
{}
NONE
ClickHouse
ClickHouse
489,471,676
5,197
alexey-milovidov
Does it reproduce on master? (you can download any green build here: https://github.com/yandex/ClickHouse/commits/master) Even if it isn't - maybe it is still make sense to add a test?
2019-05-05T23:08:44
2019-05-05T23:08:44
{}
MEMBER
ClickHouse
ClickHouse
489,636,598
5,205
den-crane
Ext. Dictionaries do not support Nulls. Just use assumeNotNull(ip) .
2019-05-06T14:15:39
2019-05-06T14:15:39
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,824,923
5,199
den-crane
What resharding? CH does not have resharding.
2019-05-06T23:35:31
2019-05-06T23:35:31
{}
CONTRIBUTOR
ClickHouse
ClickHouse
488,753,489
4,889
SaltTan
A workaround to get Null from an empty string: ``` SELECT a AS string, if(a = '', cast(NULL, 'Nullable(Decimal128(10))'), cast(concat(a, ' '), 'Decimal128(10)')) AS decimal FROM ( SELECT arrayJoin(['', ' ', '-5.666667', '5.666667', '-555555555555555555555555555.6666679999', '0', '-0', '-0.0']) AS a ...
2019-05-02T17:07:28
2019-05-02T17:07:28
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,206,485
5,185
ghost
@den-crane Thank you for the reply! > Distributed table requests 2 replicas and asks for 50% samples from one replica and other 50% from another. Yes, that's what I'd expect. However, I'm curious about the exact algorithm that is used to get 50% of samples from one replica and 50% from another one and its possibl...
2019-05-03T19:05:55
2019-05-03T19:12:22
{}
NONE
ClickHouse
ClickHouse
489,438,319
5,192
alexey-milovidov
Could you please add the performance test from Denny Crane: https://github.com/yandex/ClickHouse/issues/5190#issuecomment-489431540 (into `tests/performance`) in next PR?
2019-05-05T15:49:16
2019-05-05T15:49:16
{}
MEMBER
ClickHouse
ClickHouse
489,458,105
5,187
filimonov
That implementation does not extend supported datatime range, and make nanoseconds a 'default choice' for users who need (let's say) milliseconds only. In my opinion: 1) precision should be optionated (like in Decimal datatype), because nanoseconds are needed quite rare (while micro-/milli- seconds - quite often)....
2019-05-05T19:55:28
2019-05-05T20:17:20
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,459,844
4,860
aadant
I was replying to your question : >How do you deal with those cases in other >databases? Please implement sub-second precision (milli and micro seconds) rather than second only. That’s enough for most cases. Timestamp(9) might be possible to implement with 64 bits but I doubt Datetime(9) is doable. That is low pr...
2019-05-05T20:17:36
2019-05-05T20:17:36
{}
NONE
ClickHouse
ClickHouse
489,459,869
5,187
alexey-milovidov
> precision should be optionated (like in Decimal datatype), because nanoseconds are needed quite rare (while micro-/milli- seconds - quite often). Even milliseconds forces to use at least 64bit data type instead of 32bit. Consequently, there is not much difference between milliseconds and nanoseconds resolution. ...
2019-05-05T20:17:54
2019-05-05T20:19:02
{}
MEMBER
ClickHouse
ClickHouse
489,463,832
4,860
filimonov
> I was replying to your question. Question was about nanoseconds :) Nevermind. I glad that we agree that microseconds are needed, while nanoseconds - sounds a bit like overkill.
2019-05-05T21:12:28
2019-05-05T21:12:28
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,779,065
5,201
filimonov
Created feature issue, please see #5207
2019-05-06T21:13:59
2019-05-06T21:13:59
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,785,370
4,026
filimonov
@wawanawna Did you find the reason?
2019-05-06T21:27:01
2019-05-06T21:27:01
{}
CONTRIBUTOR
ClickHouse
ClickHouse
489,822,556
2,551
filimonov
Related #3997
2019-05-06T23:23:44
2019-05-06T23:23:44
{}
CONTRIBUTOR