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
519,598,813
6,335
filimonov
if the serverA is configured as a part of clusterX it's quite logical that serverA should also know that he's a part if that cluster, isn't it?
2019-08-08T16:47:21
2019-08-08T16:47:21
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,670,321
6,325
alexey-milovidov
Probably because `-Werror` and extra warnings are enabled only for the code inside `dbms`?
2019-08-08T20:13:18
2019-08-08T20:13:18
{}
MEMBER
ClickHouse
ClickHouse
519,823,500
6,366
alesapin
Awesome, now Stress test caught race on `changed` flag.
2019-08-09T08:05:38
2019-08-09T08:05:38
{}
MEMBER
ClickHouse
ClickHouse
519,950,995
6,400
abyss7
Should finish #5286
2019-08-09T14:56:30
2019-08-09T14:56:30
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,175,534
6,420
alexey-milovidov
It's unconvenient, because we have to set alt stack for every thread, and it won't give us much advantages.
2019-08-10T19:55:51
2019-08-10T19:55:51
{}
MEMBER
ClickHouse
ClickHouse
520,179,670
6,431
alexey-milovidov
Looks like it doesn't work for scalar subqueries, IN subqueries and JOINs?
2019-08-10T21:06:56
2019-08-10T21:06:56
{}
MEMBER
ClickHouse
ClickHouse
520,184,803
6,434
alexey-milovidov
We use re2 syntax that is a limited version of PCRE: https://clickhouse.yandex/docs/en/query_language/functions/string_search_functions/#match-haystack-pattern
2019-08-10T22:42:10
2019-08-10T22:42:23
{}
MEMBER
ClickHouse
ClickHouse
520,186,507
6,435
alexey-milovidov
``` SELECT SearchEngineID, count(), min(EventDate), max(EventDate), avg(ResolutionWidth), sum(IsNotBounce), avgIf(SendTiming, SendTiming != -1) FROM hits_100m_single GROUP BY SearchEngineID ORDER BY count() DESC LIMIT 10 ``` Before: ``` QPS: 4.730, RPS: 472976193.985, MiB/s: 4961.718, result RPS: 47.298, resul...
2019-08-10T23:13:24
2019-08-10T23:13:24
{}
MEMBER
ClickHouse
ClickHouse
520,234,738
6,431
alexey-milovidov
The method may remain const but call a specified callback. This callback can be installed in `executeQuery`. Care should be taken because it can be called concurrently.
2019-08-11T14:56:46
2019-08-11T14:56:46
{}
MEMBER
ClickHouse
ClickHouse
520,241,579
6,372
JTCunning
@alexey-milovidov: All four of the specified mounting options are already in place. If I had to make a guess, the latency comes from GCP's Persistent Disk itself. The block storage system is able to record high marks on write iops by making concurrent calls. A single-threaded application can commit upwards to 2k write ...
2019-08-11T16:30:04
2019-08-11T16:30:04
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,599,848
6,335
astudnev
> if the serverA is configured as a part of clusterX it's quite logical that serverA should also know that he's a part if that cluster, isn't it? Yes, this is satisfied. It is in own config
2019-08-08T16:50:02
2019-08-08T16:50:02
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,661,471
6,401
alexey-milovidov
Better stack trace: ``` 2019.08.08 22:44:28.643953 [ 47 ] {72c459ed-02e9-4810-8164-96c144061c1b} <Debug> executeQuery: (from 127.0.0.1:51864) CREATE TABLE xx (`date` Date, `id` Int64, `clicks` Int64, `price` Float64, `spend` Float64) ENGINE = SummingMergeTree([price, spend]) PARTITION BY toYYYYMM(date) ORDER BY id ...
2019-08-08T19:46:05
2019-08-08T19:46:05
{}
MEMBER
ClickHouse
ClickHouse
519,988,568
6,325
alexey-milovidov
> You tell me :) I have no idea. If it is enabled inside some project by means of cmake it supposed to fail in the same fashion on my and your build machine, right? Yes. Maybe the difference is in compiler version?
2019-08-09T16:50:52
2019-08-09T16:50:52
{}
MEMBER
ClickHouse
ClickHouse
520,037,154
6,092
stavrolia
There are too many scenarios for StorageFile now, and as a consequence, we have too many bool flags in class StorageFile and non-documented ideas what fields in which cases should be used. Complexity increased gradually and this PR make it a bit worse, but I don't how to change it. Seems like refactoring of StorageFil...
2019-08-09T19:29:40
2019-08-09T19:29:40
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,163,810
6,325
kreuzerkrieg
> Everything is fine! How about adding more GCC flags? I would like to add more warning flags for more strict conformance Do you mind?
2019-08-10T16:50:01
2019-08-10T16:50:01
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,242,999
6,372
JTCunning
@filimonov: Here is an illustration of my description. In this graph, we're recording `SELECT count() AS parts FROM system.parts WHERE not active;`. I've doubled our frequency of writing at 9:18. The number of inactive parts climbs because we wait for that 7.5min inactivity marker you described, and then we start t...
2019-08-11T16:48:29
2019-08-11T16:54:02
{ "+1": 2 }
CONTRIBUTOR
ClickHouse
ClickHouse
520,246,365
6,372
alexey-milovidov
BTW, one our customer has exactly the opposite experience: - when we removing data files, there is too high IO load, and they propose something to gradually truncate files before removal.
2019-08-11T17:37:13
2019-08-11T17:37:13
{}
MEMBER
ClickHouse
ClickHouse
520,271,790
6,372
JTCunning
That was extremely quick, thank you. Since this is our current production cluster on the latest version in changelog (19.11.5.28), I'll figure out how we can test this with minimal risk and report back.
2019-08-11T23:59:48
2019-08-12T00:00:38
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,662,270
6,401
alexey-milovidov
https://github.com/yandex/ClickHouse/pull/4056
2019-08-08T19:48:34
2019-08-08T19:48:34
{}
MEMBER
ClickHouse
ClickHouse
519,956,199
6,397
alexey-milovidov
Build log looks Ok. BTW, we have bugfix releases after 19.7.3 (19.7.4, 19.7.5). We need to test them.
2019-08-09T15:11:13
2019-08-09T15:11:13
{}
MEMBER
ClickHouse
ClickHouse
520,022,276
6,425
KochetovNicolai
TODO: 00963_temporary_live_view_watch_live_timeout: ``` 2019.08.09 21:20:32.725697 [ 79 ] {fb456de5-b62b-4b39-a243-2efc034881fd} <Error> executeQuery: Code: 32, e.displayText() = DB::Exception: Attempt to read after eof (version 19.13.1.1) (from [::1]:35426) (in query: WATCH test.lv), Stack trace: 0. 0xc6c9a11 St...
2019-08-09T18:37:39
2019-08-09T18:37:39
{}
MEMBER
ClickHouse
ClickHouse
520,152,078
6,415
simpl1g
Tested this build and mv started working after having this error ``` 2019.08.09 07:23:23.875452 [ 30 ] {} <Error> void DB::StorageKafka::streamThread(): Code: 32, e.displayText() = DB::Exception: Attempt to read after eof: Cannot parse Int8 from String, because value is too short: while converting source column colum...
2019-08-10T14:11:56
2019-08-10T14:11:56
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,270,519
6,372
alexey-milovidov
I've made a change for you. But now you have to test it and report the results. Download a package here: https://clickhouse-builds.s3.yandex.net/6438/28e0e64638b2487282bb0f5de85ee530dc1e55c0/report.html (from top row - gcc9, relwithdebuginfo)
2019-08-11T23:37:20
2019-08-11T23:37:20
{}
MEMBER
ClickHouse
ClickHouse
519,917,050
6,413
nikitamikhaylov
Undefined behaviour.
2019-08-09T13:22:47
2019-08-09T13:22:47
{}
MEMBER
ClickHouse
ClickHouse
519,920,549
6,398
Enmk
Thank you for reporting. I was able to reproduce the issue and working on a fix.
2019-08-09T13:33:10
2019-08-09T13:33:10
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,973,356
6,405
filimonov
Also it's quite strange that you have DEFAULT in capital letters. ClickHouse is case-sensitive.
2019-08-09T16:00:28
2019-08-09T16:00:28
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,153,270
6,325
alexey-milovidov
Everything is fine!
2019-08-10T14:29:13
2019-08-10T14:29:13
{ "+1": 1 }
MEMBER
ClickHouse
ClickHouse
520,164,944
6,433
alexey-milovidov
We also have known stack overflows in: - condition for row-level security involving a subquery on the same table; - Merge table over another Merge tables that will make a loop; - Distributed table with local shard that points to the same Distributed table; - hierarchical dictionaries with a loop.
2019-08-10T17:05:58
2019-08-10T17:05:58
{}
MEMBER
ClickHouse
ClickHouse
520,178,064
6,433
alexey-milovidov
Everything is solved except: > hierarchical dictionaries with a loop.
2019-08-10T20:38:33
2019-08-10T20:38:33
{}
MEMBER
ClickHouse
ClickHouse
520,202,857
6,372
filimonov
Great to hear that Sentry is using ClickHouse now :) (sidenote: i've used Sentry for some petproject some time ago and it's great). Can I ask how many inserts per second per server do you have? What is insert batch size? Did you consider increasing insert batch size instead of increasing number of inserts? (usually...
2019-08-11T06:01:16
2019-08-11T06:01:45
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,234,245
6,431
alexey-milovidov
Yes, we also have demand for similar feature from another user.
2019-08-11T14:50:17
2019-08-11T14:50:17
{}
MEMBER
ClickHouse
ClickHouse
519,729,146
6,410
alexey-milovidov
No test needed.
2019-08-09T00:02:10
2019-08-09T00:02:10
{}
MEMBER
ClickHouse
ClickHouse
520,183,674
6,434
den-crane
``` select x, match(x, '2[3-4]') as match from (select arrayJoin(['12345', '224', '125432']) x) ┌─x──────┬─match─┐ │ 12345 │ 1 │ │ 224 │ 1 │ │ 125432 │ 0 │ └────────┴───────┘ ```
2019-08-10T22:21:22
2019-08-10T22:21:22
{ "+1": 4 }
CONTRIBUTOR
ClickHouse
ClickHouse
520,236,516
6,372
alexey-milovidov
How does it relate to the mounting options of the filesystem? I mean noatime, nodiratime, nobarrier, writeback...
2019-08-11T15:20:58
2019-08-11T15:20:58
{}
MEMBER
ClickHouse
ClickHouse
519,678,558
6,325
kreuzerkrieg
You tell me :) I have no idea. If it is enabled inside some project by means of cmake it supposed to fail in the same fashion on my and your build machine, right?
2019-08-08T20:37:58
2019-08-08T20:37:58
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,769,604
6,184
alanpaulkwan
https://www.dropbox.com/s/8i7uyewfp5e9vgw/wdi.sas7bdat?dl=0 World Development Indicators - public data. 400k+ rows. Let me know. ----------------------------- *Alan Kwan* HK +852-9794-9483 USA +1-626-2164899 On Thu, Aug 1, 2019 at 5:57 PM filimonov <notifications@github.com> wrote: > Can you provide some conc...
2019-08-09T04:09:20
2019-08-09T04:09:20
{}
NONE
ClickHouse
ClickHouse
519,888,289
6,412
alexey-milovidov
PS. I remember, that Hive has multiple ODBC drivers with different levels of quality.
2019-08-09T11:44:13
2019-08-09T11:44:13
{}
MEMBER
ClickHouse
ClickHouse
519,889,804
6,412
frankegoesdown
@alexey-milovidov I tried with Cloudera Hive ODBC driver: https://downloads.cloudera.com/connectors/hive_odbc_2.6.1.1001/Linux/ClouderaHiveODBC-2.6.1.1001-1.x86_64.rpm and Mapr driver http://package.mapr.com/tools/MapR-ODBC/MapR_Hive/MapRHive_odbc_2.6.1.1001/ In Mapr documentation they says: https://mapr.com/d...
2019-08-09T11:50:14
2019-08-09T12:54:42
{}
NONE
ClickHouse
ClickHouse
519,917,144
6,411
alexey-milovidov
Ok. PS. What is the use case for explicit Enum8/Enum16 data types? - you should specify Enum16 if it will have less than 256 values right now but you definitely sure that it will extend past 256 values.
2019-08-09T13:23:04
2019-08-09T13:23:04
{}
MEMBER
ClickHouse
ClickHouse
519,935,246
6,397
alexey-milovidov
BTW, Do you use Delta codec in your tables?
2019-08-09T14:13:47
2019-08-09T14:13:47
{}
MEMBER
ClickHouse
ClickHouse
520,253,519
6,074
alexey-milovidov
@akuzm Too long to wait.
2019-08-11T19:23:47
2019-08-11T19:23:47
{}
MEMBER
ClickHouse
ClickHouse
519,600,464
6,394
astudnev
However, it is not clear how to convert table to the new format?
2019-08-08T16:51:38
2019-08-08T16:51:38
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,883,376
5,126
4ertus2
There's a setting any_join_distinct_right_table_keys now that disables incorrect ANY JOINS. Default behaviour is reserved for SEMI-JOIN like ones: ANY LEFT JOIN - distinct keys in the right table (it's already OK) ANY RIGHT JOIN - distinct keys in the left table (TODO) ANY INNER | FULL JOIN - distinct keys in both ...
2019-08-09T11:23:48
2019-08-09T11:27:47
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,991,224
6,325
kreuzerkrieg
dunno, but. after minor changes and push of commit, GCC magically passed build for clang header issue with inheritance access level, looks like it is GCC bug, I've just changed the access level and looks like it is working. now rebasing on master and will push shortly
2019-08-09T16:59:05
2019-08-09T16:59:05
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,005,121
6,012
abyss7
Should implement #5834
2019-08-09T17:42:59
2019-08-09T17:42:59
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,153,216
6,325
alexey-milovidov
No, it isn't related.
2019-08-10T14:28:14
2019-08-10T14:28:14
{}
MEMBER
ClickHouse
ClickHouse
519,606,539
6,335
filimonov
Do you have 'archive_servers' configured properly and in the same way on both ch1 anc ch2? Do you use node name ch1 one some other clusters using same zookeeper?
2019-08-08T17:08:29
2019-08-08T17:08:39
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,885,057
6,412
alexey-milovidov
> It telegram channel Denny Crane said that is because CH request call for discover quotes " or ``` That's right. And Hive driver does not implement this function at all. Probably it means that we should send queries without quoting of identifiers or fallback to double quotes. That's quite easy to do.
2019-08-09T11:30:56
2019-08-09T11:32:00
{}
MEMBER
ClickHouse
ClickHouse
519,958,036
6,397
thomas955
currently we are trying the latest packages (19.13.1.11) from repository over the weekend with the same data and commands.
2019-08-09T15:16:27
2019-08-09T15:16:27
{}
NONE
ClickHouse
ClickHouse
519,958,875
6,397
alexey-milovidov
Ok, looking forward for the results...
2019-08-09T15:18:52
2019-08-09T15:18:52
{}
MEMBER
ClickHouse
ClickHouse
519,981,301
6,324
filimonov
That test now kills the server: ``` CREATE TABLE test1 (a UInt8) ENGINE MergeTree ORDER BY a; CREATE TABLE test2 (a UInt8) ENGINE MergeTree ORDER BY a; CREATE MATERIALIZED VIEW mv1 TO test1 AS SELECT a FROM test2; CREATE MATERIALIZED VIEW mv2 TO test2 AS SELECT a FROM test1; insert into test1 values (1); ``` ...
2019-08-09T16:26:11
2019-08-09T16:26:11
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,156,144
6,428
stavrolia
@den-crane, the idea of improvement of implicit casts of LowCardinality is good, but unfortunately it is very expensive because of current architecture. So, I think this change is enough for scope of this issue. Could you please create another one?
2019-08-10T15:09:45
2019-08-10T15:09:45
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,208,197
6,431
rochoa
> We can propose another implementation - to account all the tables that were actually read or written and to also write this info into query_log. Sure. Any help will be much appreciated. My PR was about discussing the feature. If you think is worth working on it, I will be more than happy to change the implement...
2019-08-11T07:54:01
2019-08-11T07:54:01
{}
NONE
ClickHouse
ClickHouse
520,234,896
6,431
alexey-milovidov
We should add this info to `query_log` in form of two parallel arrays: ``` tables.database Array(String), tables.table Array(String) ```
2019-08-11T14:59:23
2019-08-11T14:59:23
{}
MEMBER
ClickHouse
ClickHouse
519,772,169
2,161
thinhnp-it
Thanks a lot guys. I'm new beginner to ClickHouse and also confusing like @filimonov. And I have a question, how can we determine the size of a shard ? OR how many shards should we use ? I'm building a real-time event analytic for 50 millions events per day with 6 server (3 SSD and 3 HDD) Thanks again.
2019-08-09T04:25:22
2019-08-09T04:41:34
{}
NONE
ClickHouse
ClickHouse
519,920,542
6,397
filimonov
May be related? #5002 (or not, there was memory access violation with similar stacktrace there)
2019-08-09T13:33:09
2019-08-09T13:36:23
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,156,500
6,428
den-crane
OK. Done https://github.com/yandex/ClickHouse/issues/6432
2019-08-10T15:14:37
2019-08-10T15:14:37
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,168,038
6,432
alexey-milovidov
That's exactly as it was designed.
2019-08-10T17:53:31
2019-08-10T17:53:31
{}
MEMBER
ClickHouse
ClickHouse
519,689,717
6,325
kreuzerkrieg
BTW, with CLang I get following >> error: cannot cast private base class 'std::__detail::__variant::_Variant_storage<false, std::unique_ptr<DB::SortedLookupVector<DB::AsofRowRefs::Entry<unsigned int>, unsigned int>, std::default_delete<DB::SortedLookupVector<DB::AsofRowRefs::Entry<unsigned int>, unsigned int> > >, std...
2019-08-08T21:13:42
2019-08-08T21:13:42
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,920,823
6,413
alexey-milovidov
Ok. At least this implementation will lower probability of unrecoverable state.
2019-08-09T13:33:55
2019-08-09T13:33:55
{}
MEMBER
ClickHouse
ClickHouse
519,933,868
6,397
thomas955
Complete build log is attached. We didnt try to reproduce this with the latest version yet. I will try this over the weekend with the actual release version from the Bionic repository. [build_log_clickhouse.log](https://github.com/yandex/ClickHouse/files/3486701/build_log_clickhouse.log)
2019-08-09T14:09:58
2019-08-09T14:09:58
{}
NONE
ClickHouse
ClickHouse
519,999,912
6,341
den-crane
Check something simple curl 'localhost:8123/' Ok. curl 'localhost:8123/?query=select%201' 1 curl 'localhost:8123/?query=select%201%20x%20where%20x' 1
2019-08-09T17:26:46
2019-08-09T17:27:36
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,234,152
6,436
alexey-milovidov
Ok, it's pretty easy to add this metric to ProfileEvents. PS. As an alternative way to monitor count of merges, you can enable `part_log`. Look at `<part_log>` in configuration file.
2019-08-11T14:49:01
2019-08-11T14:49:01
{}
MEMBER
ClickHouse
ClickHouse
519,916,623
6,413
alexey-milovidov
Ok. > If an error occured, we rename metadata back. And what if we have partially removed some ZK data before the error happened?
2019-08-09T13:21:25
2019-08-09T13:21:25
{}
MEMBER
ClickHouse
ClickHouse
520,109,944
6,428
den-crane
The main issue is not the "allow_suspicious_low_cardinality_types" but LowCardinality itself. ``` desc (select x='a' bool from (select toLowCardinality('a') x)) ┌─name─┬─type──────────────────┬ │ bool │ LowCardinality(UInt8) │ └──────┴───────────────────────┴ ``` EXPECTED: ``` ┌─name─┬─type──┬ │ bool │ UI...
2019-08-10T02:15:25
2019-08-10T02:16:33
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,169,038
5,967
alexey-milovidov
You should also install at least one of: - libc++ (this is clang's native implementation of C++ standard library); - recent version of g++ (it will also bring recent version of libstdc++ that clang can also use).
2019-08-10T18:10:09
2019-08-10T18:10:09
{}
MEMBER
ClickHouse
ClickHouse
520,175,562
6,420
alexey-milovidov
Superseded by https://github.com/yandex/ClickHouse/pull/6433
2019-08-10T19:56:12
2019-08-10T19:56:12
{}
MEMBER
ClickHouse
ClickHouse
520,176,942
6,432
alexey-milovidov
It is worse in storage but it may give advantages in calculations. For example, we can write something like: `low_cardinality = 'a' ? 'Hello' : 'World'` and we will get `LowCardinality(String)` and it will be much better than `String`, because Strings Hello and World will be stored just once in a dictionary. ...
2019-08-10T20:18:42
2019-08-10T20:25:17
{}
MEMBER
ClickHouse
ClickHouse
520,186,168
6,435
alexey-milovidov
In extreme case it improves performance about 1.5 times: ``` SELECT SearchEngineID, count(), min(EventDate), max(EventDate) FROM hits_100m_single GROUP BY SearchEngineID ORDER BY count() DESC LIMIT 10 ``` Before: 0.100 sec. After: 0.061 sec.
2019-08-10T23:08:47
2019-08-10T23:08:47
{}
MEMBER
ClickHouse
ClickHouse
519,808,212
6,393
Chocorean
`clickhouse-client --query "select * from system.merges"` ``` default credential 1.143259521 0 1 ['20190729_28016_791975_4_821925'] 20190729_28016_791975_4_1231304 20190729 1 11172356 86 0 0 0 0 0 567393504 12 default credential 0.952820128 0 1 ['20190805_813035_823186_4_872908'] 20190805_813035_823186_4_1231304 201...
2019-08-09T07:14:46
2019-08-09T07:16:06
{}
NONE
ClickHouse
ClickHouse
519,927,350
6,397
alexey-milovidov
This is related to alternative build. Because we don't use `Poco::SignalHandler::handleSignal(int)` in our code (even in implicit way). Most likely you have used libraries from the OS packages (that is discouraged). We need information about: - full build log; - does the issue reproduce if you install the latest...
2019-08-09T13:51:57
2019-08-09T13:51:57
{}
MEMBER
ClickHouse
ClickHouse
519,972,879
6,405
filimonov
> There is no table DEFAULT.test_local on server: db2:9000 Are you sure you have such a table there?
2019-08-09T15:59:03
2019-08-09T15:59:03
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,123,859
6,325
kreuzerkrieg
@alexey-milovidov it looks better now could you please check the performance test and let me know if failures related to my changes. If yes I think I will need some guidance from you
2019-08-10T06:50:00
2019-08-10T06:50:00
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,170,439
6,432
den-crane
OK. But as I remember LowCardinality(UInt8) is always worse than UInt8?
2019-08-10T18:30:37
2019-08-10T18:30:37
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,179,784
6,430
abyss7
Does server have any errors in logs related to columns' casting?
2019-08-10T21:08:49
2019-08-10T21:08:49
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,187,475
6,435
alexey-milovidov
For the case without key, the performance advantage is even more obvious (4.5 times): ``` SELECT count(), min(EventDate), max(EventDate) FROM hits_100m_single QPS: 13.025, RPS: 1302511482.858, MiB/s: 2484.343, result RPS: 13.025, result MiB/s: 0.000. 0.000% 0.067 sec. 10.000% 0.070 sec. 20.000% 0.071 sec. 3...
2019-08-10T23:31:11
2019-08-10T23:31:11
{}
MEMBER
ClickHouse
ClickHouse
519,653,556
6,394
alesapin
It will be converted automatically after merges.
2019-08-08T19:21:22
2019-08-08T19:21:22
{}
MEMBER
ClickHouse
ClickHouse
519,953,574
5,128
filimonov
I've found yesterday that rdkafka have statistics API: https://github.com/edenhill/librdkafka/blob/master/STATISTICS.md Also cppkafka has wrapper for that: https://github.com/mfontanini/cppkafka/blob/29cb02b756a79a9f77c36e95cbcf799d3e3edfc2/src/configuration.cpp#L90 That can be exposed in something like syst...
2019-08-09T15:03:31
2019-08-09T15:03:31
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,149,238
6,393
den-crane
You can cancel mutations by kill mutation or manually. You can reduce background_pool_size / background_schedule_pool_size -- number of simultaneously running tasks.
2019-08-10T13:32:03
2019-08-10T13:32:03
{ "+1": 1 }
CONTRIBUTOR
ClickHouse
ClickHouse
520,163,496
6,433
alexey-milovidov
TODO: ``` perl -e 'print "SELECT 1" . ("+1" x 10000) ' | curl 'http://localhost:8123/' -d @- ```
2019-08-10T16:45:30
2019-08-10T16:45:30
{}
MEMBER
ClickHouse
ClickHouse
520,194,441
6,399
alexey-milovidov
Let's merge with master, because we have fixes for integration and performance tests.
2019-08-11T02:19:08
2019-08-11T02:19:08
{}
MEMBER
ClickHouse
ClickHouse
520,202,012
1,424
hodgesrm
@alexey-milovidov Is there a PR for this yet? Thx.
2019-08-11T05:41:10
2019-08-11T05:41:10
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,204,183
6,372
JTCunning
Sure! Here is our breakdown: Main cluster is 3 shards with 3 replicas. Inserts are roughly once a second in order to reduce latency in actions triggered off of individual rows that mean conditions ([alerts with time intervals](https://docs.sentry.io/workflow/notifications/alerts/)). We have considered increasing bat...
2019-08-11T06:31:59
2019-08-11T06:32:09
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,234,981
1,424
alexey-milovidov
No, it is in initial stage of development.
2019-08-11T15:00:34
2019-08-11T15:00:34
{}
MEMBER
ClickHouse
ClickHouse
520,236,893
6,372
alexey-milovidov
Parallel files removal is not a hard thing to do... PS. Do you use enough recent version? (recently we have made a small optimization in this place): https://github.com/yandex/ClickHouse/blob/master/CHANGELOG.md#performance-improvement-1
2019-08-11T15:26:02
2019-08-11T15:26:02
{}
MEMBER
ClickHouse
ClickHouse
520,261,096
6,430
josepowera
Nope, unfortunatelly nothing regarding casting.... No useful error either. Just thinking... We have 60 tables, and "BackgroundProcessingPool with 16 threads". Could this be somehow related? Here are real table / view for our kafka-mv-table ``` biddetaileventdata_queueraw -> mv biddetaileventdata_MV_queueraw...
2019-08-11T21:16:10
2019-08-11T21:16:10
{}
NONE
ClickHouse
ClickHouse
519,662,361
6,366
alesapin
Atomic settings seem slightly ugly, but I don't see better solutions without huge refactoring of setting usage in MergeTree.
2019-08-08T19:48:49
2019-08-08T19:48:49
{}
MEMBER
ClickHouse
ClickHouse
519,669,616
6,325
kreuzerkrieg
@alexey-milovidov just ran the following line >> cmake .. -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_VERBOSE_MAKEFILE=ON -DCMAKE_BUILD_TYPE=None -DCMAKE_INSTALL_SYSCONFDIR=/etc -DCMAKE_INSTALL_LOCALSTATEDIR=/var -DCMAKE_EXPORT_NO_PACKAGE_REGISTRY=ON -DCMAKE_FIND_PACKAGE_NO_PACKAGE_REGISTRY=ON -DSANITIZE= -DENABLE_EMBEDDED_CO...
2019-08-08T20:11:04
2019-08-08T20:11:04
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,942,456
6,397
thomas955
No.
2019-08-09T14:33:42
2019-08-09T14:33:42
{}
NONE
ClickHouse
ClickHouse
519,954,100
6,403
filimonov
Also: https://github.com/yandex/ClickHouse/issues/5749
2019-08-09T15:05:00
2019-08-09T15:05:00
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,096,104
6,430
filimonov
Maybe related https://github.com/yandex/ClickHouse/pull/6415
2019-08-09T23:39:09
2019-08-09T23:39:09
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,168,566
6,324
alexey-milovidov
https://github.com/yandex/ClickHouse/pull/6433
2019-08-10T18:01:50
2019-08-10T18:01:50
{}
MEMBER
ClickHouse
ClickHouse
520,179,861
6,431
alexey-milovidov
We can propose another implementation - to account all the tables that were actually read or written and to also write this info into query_log.
2019-08-10T21:09:59
2019-08-10T21:09:59
{}
MEMBER
ClickHouse
ClickHouse
520,207,407
5,286
abyss7
Resolved in v19.14+
2019-08-11T07:39:19
2019-08-11T07:39:19
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,234,584
6,431
alexey-milovidov
Maybe we can hook it (accounting of used tables) somewhere in `Context::getTableImpl` method. It should work for all type of queries including CREATE, DROP... Note that this is the same place where access to databases is checked. Looks like a natural place.
2019-08-11T14:54:52
2019-08-11T14:57:40
{}
MEMBER
ClickHouse
ClickHouse
520,275,473
6,372
JTCunning
From my initial testing on a single machine we use for access logs, this seems to be working great. I can see multiple threads in the `BackgrProcPool` pinning IO in iotop. The ioutil graph is pinned at 100% and the iowait graph is going higher than a single core's percentage. More importantly, the obsoleted parts ar...
2019-08-12T00:46:44
2019-08-12T00:46:44
{}
CONTRIBUTOR
ClickHouse
ClickHouse
519,957,927
6,418
den-crane
probably it will be useful for at least for SYSTEM DROP DNS CACHE; SYSTEM DROP MARKS CACHE; SYSTEM RELOAD CONFIG;
2019-08-09T15:16:07
2019-08-09T15:16:07
{}
CONTRIBUTOR
ClickHouse
ClickHouse
520,177,403
6,325
alexey-milovidov
We need to add as many warnings as possible.
2019-08-10T20:26:20
2019-08-10T20:26:20
{}
MEMBER
ClickHouse
ClickHouse
520,233,268
6,435
alexey-milovidov
Performance improvement is obvious from performance test.
2019-08-11T14:36:50
2019-08-11T14:36:50
{}
MEMBER
ClickHouse
ClickHouse
523,142,087
215
filimonov
So It's actually another issue. Human-readable format is accepted. If you think that numeric form shoul be also accepted - please create another issue with motivation. BTW, the workaround is simple - use `CAST(1,'Enum8(\'TRUE\' = 0, \'FALSE\' = 1')`
2019-08-20T18:35:27
2019-08-20T18:36:20
{}
CONTRIBUTOR