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 | 534,056,770 | 7,057 | zhang2014 | @alexey-milovidov Can you add test label to this pr? This pr is not compile in my environment and I'm not sure if the error is caused by my environment configuration | 2019-09-23T11:18:43 | 2019-09-23T11:18:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,070,193 | 5,907 | nicelulu | https://github.com/ClickHouse/ClickHouse/blob/7a822ad815f725f275202f57f15b541d9f73e8ac/dbms/src/Interpreters/InterpreterSelectQuery.cpp#L1498-L1499
Is this a similar optimization? | 2019-09-23T12:06:41 | 2019-09-23T12:06:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,082,843 | 7,037 | abyss7 | That merge was very unexpected. I could just take missing changes to my PR. | 2019-09-23T12:46:09 | 2019-09-23T12:46:09 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,252,967 | 7,026 | akuzm | @amosbird @alexey-milovidov @KochetovNicolai for your consideration | 2019-09-23T19:41:04 | 2019-09-23T19:41:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,527,757 | 6,981 | proller | > But it is the only right way, because production version of ClickHouse can be build only with specific versions of specific libraries.
... In specific ubuntu 18.04 x86_64, but there are dozens of operating systems and architectures in the world | 2019-09-24T12:06:55 | 2019-09-24T12:06:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,558,665 | 7,078 | silviucpp | @alexey-milovidov my feeeling is that `isSuitableForConstantFolding` is not working properly. Problem is that function is executed several times during the query analysis. | 2019-09-24T13:30:04 | 2019-09-24T13:30:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 533,884,177 | 7,009 | dujijun007 | > Interesting. Can you show create table statement? Engine, Keys (order by, partition by, skipping index) , and columns used in query should be enough.
>
> Also please repeat both queries in clickhouse-client, with `set send_logs_level='trace' `and send the results.
CREATE TABLE datasets.user_event_log4 (`event_i... | 2019-09-22T13:57:22 | 2019-09-22T13:57:22 | {} | NONE |
ClickHouse | ClickHouse | 534,120,127 | 6,892 | akuzm | Should update to 1.9.4: https://pocoproject.org/blog/?p=1099 | 2019-09-23T14:19:08 | 2019-09-23T14:19:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,212,606 | 6,975 | alexey-milovidov | Ok. It's not an error - just inflated memory usage due to TSan instrumentation.
(memory is exhausted before stack overflow check happens)
Let's disable this test under TSan (this is done in sandbox scripts in Arcadia). | 2019-09-23T17:56:27 | 2019-09-23T17:57:04 | {} | MEMBER |
ClickHouse | ClickHouse | 534,525,917 | 6,981 | proller | > We can use single UNBUNDLED option.
cmake -D UNBUNDLED=1 ..
Its broken. Now you should manually disable USE_SNAPPY and soon all libs not installed in system.
For minimal build you should specify ~30 options. | 2019-09-24T12:01:11 | 2019-09-24T12:01:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,561,505 | 7,065 | abyss7 | @amosbird Can you explain in details what is this fix about and why the issue can't be resolved with typical CMake dependencies and interfaces? Overriding default macro looks very hacky, generally it may lead to new problems, when someone decides to do any refactoring months later. | 2019-09-24T13:36:43 | 2019-09-24T13:36:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,564,762 | 7,065 | amosbird | @abyss7 When doing shared builds, TUs which try to override std symbols (such as malloc or operator new) should be statically linked into the final executable and only be linked once. It includes all unit testing executables. I cannot find easy way of collecting executables in one place other than overriding default ma... | 2019-09-24T13:43:57 | 2019-09-24T13:46:07 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 534,815,373 | 6,984 | ghost | @zhang2014
I appreciate your code review.
It looks good idea extending this method to other skip indices.
**In the future**, it is possible to contribute the support of nested array type. :) | 2019-09-25T01:47:55 | 2019-09-25T01:55:29 | {} | NONE |
ClickHouse | ClickHouse | 534,051,191 | 5,237 | yingfeng | @Onlinehead
Has this issue been fixed right now? It seems my [situation](https://github.com/yandex/ClickHouse/issues/7054) has some similarities with yours except that the IN query is replaced with a scalar subquery.
| 2019-09-23T10:58:35 | 2019-09-23T10:58:35 | {} | NONE |
ClickHouse | ClickHouse | 534,151,469 | 6,817 | amosbird | @alesapin Does this still happen in current HEAD ? | 2019-09-23T15:27:52 | 2019-09-23T15:27:52 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 534,223,731 | 7,066 | alexey-milovidov | > -fno-optimize-sibling-calls
That requires clarification. | 2019-09-23T18:24:49 | 2019-09-23T18:24:49 | {} | MEMBER |
ClickHouse | ClickHouse | 534,241,092 | 7,067 | reflection | ```
CREATE TABLE IF NOT EXISTS wsg.wsg_wf (
timestamp DateTime,
sn LowCardinality(Nullable(String)),
action Nullable(UInt8),
app_rule LowCardinality(Nullable(String)),
category Array(LowCardinality(Nullable(String))),
content_type LowCardinalit... | 2019-09-23T19:09:06 | 2019-09-23T19:09:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,244,852 | 6,975 | Akazz | > (memory is exhausted before stack overflow check happens)
Exactly!
I've added this 00985_merge_stack_overflow to the set of tests to be skipped under TSan. Hopefully this relieves the annoyance until we find a better solution (perhaps via native handler behind the std::thread) | 2019-09-23T19:18:54 | 2019-09-23T19:18:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,257,901 | 7,014 | alexey-milovidov | It's merged automatically. | 2019-09-23T19:54:34 | 2019-09-23T19:54:34 | {} | MEMBER |
ClickHouse | ClickHouse | 534,444,775 | 508 | cailin-lai | According to my experiences, this issue typically indicates the configuration is incorrect for node "interserver_http_host" in config.xml. ;-) | 2019-09-24T08:10:20 | 2019-09-24T08:10:20 | {} | NONE |
ClickHouse | ClickHouse | 534,483,927 | 4,689 | georgeakulov | Receive similiar error:
019.09.24 11:42:45.607806 [ 230 ] {61e183ad-49a8-4bd8-aa15-7178a2a0d127} <Error> virtual DB::WriteBufferFromHTTPServerResponse::~WriteBufferFromHTTPServerResponse(): Poco::Except
ion. Code: 1000, e.code() = 0, e.displayText() = Assertion violation: !_pStream in file "/var/tmp/portage/dev-libs/... | 2019-09-24T09:52:10 | 2019-09-24T09:52:10 | {} | NONE |
ClickHouse | ClickHouse | 534,512,520 | 7,077 | alexey-milovidov | Could you please add a performance test to `dbms/tests/performance`?
The `default.trips_mergetree` is available in our CI, look at the example in `nyc_taxi.xml`.
Then we will make some tricks and try to improve performance. | 2019-09-24T11:19:00 | 2019-09-24T11:19:00 | {} | MEMBER |
ClickHouse | ClickHouse | 534,568,746 | 7,078 | alexey-milovidov | Yes. Now we can find all locations with gdb or by printing a stack trace. | 2019-09-24T13:53:09 | 2019-09-24T13:53:15 | {} | MEMBER |
ClickHouse | ClickHouse | 534,569,904 | 7,007 | alexey-milovidov | For short-circuit constant folding in logical expressions, look at https://github.com/yandex/ClickHouse/pull/4672 (this PR was unfinished but we want to continue). | 2019-09-24T13:55:46 | 2019-09-24T13:55:46 | {} | MEMBER |
ClickHouse | ClickHouse | 533,928,552 | 7,049 | alexey-milovidov | A query like
`SELECT repeat('', 1000000000)`
opens the possibility for DoS. | 2019-09-22T23:34:10 | 2019-09-22T23:34:10 | {} | MEMBER |
ClickHouse | ClickHouse | 534,145,688 | 6,975 | Akazz | clickhouse-server crash looks like this under GDB:
```
Program terminated with signal SIGKILL, Killed.
The program no longer exists.
(gdb)
```
And the reason turned out to be pretty commonplace - OOM:
```
[Mon Sep 23 17:51:35 2019] Out of memory: Kill process 48578 (clickhouse-serv) score 939 or sacrifice chi... | 2019-09-23T15:15:52 | 2019-09-23T15:15:52 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 533,926,484 | 6,969 | azat | >70bc895
great, thanks! | 2019-09-22T23:00:09 | 2019-09-22T23:00:09 | {} | MEMBER |
ClickHouse | ClickHouse | 534,090,104 | 7,037 | abyss7 | Anyway @silviucpp thanks for the nano-seconds sleep and pthread stack-size solutions! | 2019-09-23T13:06:01 | 2019-09-23T13:06:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,223,360 | 7,066 | alexey-milovidov | > It doesn't link in Debug, because executables are too big
We can fallback to `gold` if it is `lld` issue. | 2019-09-23T18:23:54 | 2019-09-23T18:23:54 | {} | MEMBER |
ClickHouse | ClickHouse | 534,255,164 | 7,071 | alexey-milovidov | It's not trivial. The table may require implicit filtering and will read more rows.
The only way is to analyze limit inside IStorage::read (for MergeTree and Numbers storages) and adapt the predicted number of rows to read. | 2019-09-23T19:47:05 | 2019-09-23T19:47:05 | {} | MEMBER |
ClickHouse | ClickHouse | 534,321,129 | 6,945 | alexey-milovidov | In debug build:
```
no space left on device: /var/lib/clickhouse/data/default/table
2019-09-23 22:53:27 Total space: 3.44 TiB
2019-09-23 22:53:27 Available space: 0.00 B
2019-09-23 22:53:27 Total inodes: 234.30 million
2019-09-23 22:53:27 Available inodes: 222.05 million
2019-09-23 22:53:27 Mount point: /
201... | 2019-09-23T23:16:06 | 2019-09-23T23:16:06 | {} | MEMBER |
ClickHouse | ClickHouse | 534,520,094 | 6,835 | alexey-milovidov | Ok. | 2019-09-24T11:43:08 | 2019-09-24T11:43:08 | {} | MEMBER |
ClickHouse | ClickHouse | 534,527,244 | 7,080 | filimonov | Typical usecase: server was off for a long time / replace replica with new one.
Currently it will not allow you to CREATE (replica already exists), ATTACH has not to much sense and will requite enabling replica recovery, and will try to 'execute' obsolete queue. | 2019-09-24T12:05:16 | 2019-09-24T12:05:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,720,400 | 7,026 | akuzm | Looks better with inlining:
```
SELECT URL, count() AS c FROM hits_100m_single GROUP BY URL ORDER BY c DESC LIMIT 10
┏━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━━━┓
┃ version ┃ count() ┃ min ┃ med ┃ max ┃ q95-q05 ┃
┡━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━╇━━━━━━━╇━━━━━... | 2019-09-24T19:50:46 | 2019-09-24T19:50:46 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 533,922,329 | 7,007 | alexey-milovidov | > Maybe should not be doing CreatingSetsBlockInputStream return quickly?
Yes, this is not the correct way of implementation.
We need to check if the set is empty deeper in query execution pipeline. | 2019-09-22T22:00:31 | 2019-09-22T22:00:31 | {} | MEMBER |
ClickHouse | ClickHouse | 533,961,124 | 6,984 | ghost | @zhang2014
Thank you for your comments.
I'm felling that your insight for CH is great.
I fixed source codes. | 2019-09-23T05:06:59 | 2019-09-23T05:06:59 | {} | NONE |
ClickHouse | ClickHouse | 533,966,831 | 6,984 | zhang2014 | 👍 The code looks great.
BTW: Array(Array) currently does not appear to be supported. maybe we could consider using Array(bloom_filter_bits) to save array bloom filter?this way we will support nesting of any number of Array types and the storage costs will always be the same. because:`(array_row_1_size + array_row_... | 2019-09-23T05:48:20 | 2019-09-23T05:48:20 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,083,888 | 5,601 | vasyaabr | +1
It would be efficient to store queues in Avro format, because of smaller size. | 2019-09-23T12:49:17 | 2019-09-23T12:49:17 | {} | NONE |
ClickHouse | ClickHouse | 534,194,418 | 6,817 | alesapin | @amosbird found that error persists only with `clang` compiler, for `gcc-9` it's already fixed. | 2019-09-23T17:11:10 | 2019-09-23T17:11:10 | {} | MEMBER |
ClickHouse | ClickHouse | 534,259,615 | 7,072 | alexey-milovidov | 512 KB stack is Ok. It isn't much worse than Linux default 2 MB and everything should work correctly. | 2019-09-23T19:59:14 | 2019-09-23T19:59:14 | {} | MEMBER |
ClickHouse | ClickHouse | 534,320,361 | 7,066 | alexey-milovidov | Ok, but questions remain. | 2019-09-23T23:12:39 | 2019-09-23T23:12:39 | {} | MEMBER |
ClickHouse | ClickHouse | 534,520,600 | 6,981 | alexey-milovidov | > Without MISSING_ logic to build without submodules tens of options should be explicitly specified.
We can use single UNBUNDLED option.
`cmake -D UNBUNDLED=1 ..` | 2019-09-24T11:44:43 | 2019-09-24T11:44:43 | {} | MEMBER |
ClickHouse | ClickHouse | 534,783,708 | 6,927 | alexey-milovidov | Not descriptive enough. I've edited the text a little. | 2019-09-24T23:14:50 | 2019-09-24T23:14:50 | {} | MEMBER |
ClickHouse | ClickHouse | 534,223,019 | 7,066 | alexey-milovidov | Why `-O1`? It may lead to missed bugs. | 2019-09-23T18:23:05 | 2019-09-23T18:23:05 | {} | MEMBER |
ClickHouse | ClickHouse | 534,227,122 | 6,912 | alexey-milovidov | Let's propose some different names:
`fillSortedArraysCumulative` | 2019-09-23T18:33:12 | 2019-09-23T18:33:12 | {} | MEMBER |
ClickHouse | ClickHouse | 534,230,738 | 6,912 | alexey-milovidov | Maybe you need two functions instead:
```
SELECT mapFill([1, 2, 6], [10, 20, 30])
([1, 2, 3, 4, 5, 6], [10, 20, 0, 0, 0, 30])
```
```
SELECT arrayCumSum([10, 20, 0, 0, 0, 30])
[10, 30, 30, 30, 30, 60]
``` | 2019-09-23T18:42:30 | 2019-09-23T18:42:30 | {} | MEMBER |
ClickHouse | ClickHouse | 534,781,422 | 7,086 | alexey-milovidov | Debug tests: `No space left on device` (CI issue). | 2019-09-24T23:04:47 | 2019-09-24T23:04:47 | {} | MEMBER |
ClickHouse | ClickHouse | 534,284,227 | 7,072 | silviucpp | Yes so far I don't find any issue with this patch. I'll run some more tests tomorrow. | 2019-09-23T21:05:19 | 2019-09-23T21:05:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,322,777 | 7,028 | alexey-milovidov | Up next:
- a command to download SDK in build instruction;
- add to CI and provide tarball to download. | 2019-09-23T23:23:08 | 2019-09-23T23:23:08 | {} | MEMBER |
ClickHouse | ClickHouse | 534,519,070 | 6,981 | proller | > We should provide source tarballs that are ready to build.
Beforre https://github.com/ClickHouse/ClickHouse/commit/b4339f266df17c0cd873ef10249647409e66f229 github's .zip without submodules with <10M size was enough for build.
If you want include contrib to src - it will be not used for packaging and src size wi... | 2019-09-24T11:39:47 | 2019-09-24T11:39:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,621,268 | 7,084 | den-crane | Wrong repo. try https://github.com/ClickHouse/clickhouse-odbc | 2019-09-24T15:47:44 | 2019-09-24T15:47:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,073,414 | 6,817 | alesapin | I have got slightly different trace under gdb:
```
#0 0x000000000022d805 in extent_heap_remove (ph=0x7fffe8207fd8, phn=0x7fffe8a0c500) at ../contrib/jemalloc/src/extent.c:289
#1 0x0000000000226972 in arena_bin_slabs_nonfull_remove (slab=<optimized out>, bin=<optimized out>) at ../contrib/jemalloc/src/arena.c:957
... | 2019-09-23T12:16:35 | 2019-09-23T12:16:35 | {} | MEMBER |
ClickHouse | ClickHouse | 534,250,950 | 7,026 | akuzm | This has some easy conflicts with master due to recent sparse dictionary changes. It can be reviewed and tested this way, I'll rebase later. | 2019-09-23T19:36:03 | 2019-09-23T19:36:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,511,765 | 7,077 | alexey-milovidov | The reason is escaping of zero bytes by `\0`.
Serialized aggregation state has multiple zero bytes and all of them are escaped. This is slow path of string escaping function.
> Why String FieldVisitorToString::operator() (const AggregateFunctionStateData & x) const is called so many times for the scalar sub query?
... | 2019-09-24T11:16:25 | 2019-09-24T11:16:43 | {} | MEMBER |
ClickHouse | ClickHouse | 534,565,988 | 7,083 | ghost | Thank you for explanation! | 2019-09-24T13:46:53 | 2019-09-24T13:46:53 | {} | NONE |
ClickHouse | ClickHouse | 534,568,477 | 7,065 | abyss7 | > @abyss7 When doing shared builds, TUs which try to override std symbols (such as malloc or operator new) should be directly linked into the final executable and only be linked once. It includes all unit testing executables. I cannot find easy way of collecting executables in one place other than overriding default ma... | 2019-09-24T13:52:33 | 2019-09-24T13:53:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,714,976 | 7,065 | amosbird | Tbh, I still don't have a clear mind of all other broken cases, but the way in current PR should work. I've asked for a minimal broken example in https://github.com/jemalloc/jemalloc/issues/1636 . Hopefully we can get a better idea of what might go wrong when linking to a DSO containing jemalloc symbols. | 2019-09-24T19:34:57 | 2019-09-24T19:34:57 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 533,922,144 | 7,007 | alexey-milovidov | @nicelulu That becomes more interesting.
To get proper result for aggregation query, you have to provide empty data stream before Aggregating and not after CreatingSets. Let's think how to do it... | 2019-09-22T21:57:48 | 2019-09-22T21:57:48 | {} | MEMBER |
ClickHouse | ClickHouse | 533,930,223 | 6,999 | alexey-milovidov | Now it's ready for merge!
I've made a few edits in https://github.com/yandex/ClickHouse/pull/7048 | 2019-09-22T23:57:26 | 2019-09-22T23:57:26 | {} | MEMBER |
ClickHouse | ClickHouse | 534,155,413 | 6,981 | abyss7 | As it was previously discussed in the private dev group that unintended or implicit switch from internal libraries to external ones is a very bad thing. If you want to build without submodules, then explicitly specify `-DUNBUNDLED=ON` or specify options for each component that you want to turn on-off. By default everyt... | 2019-09-23T15:36:37 | 2019-09-23T15:36:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,200,783 | 6,817 | amosbird | fixed in https://github.com/ClickHouse/ClickHouse/pull/7065 | 2019-09-23T17:27:15 | 2019-09-23T17:27:15 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 534,225,237 | 6,912 | alexey-milovidov | One test has failed.
> Functional stateless tests (ubsan) — fail:21, passed:944, skipped:2
And this is unrelated, it is fixed in master. | 2019-09-23T18:28:34 | 2019-09-23T18:28:34 | {} | MEMBER |
ClickHouse | ClickHouse | 534,451,003 | 7,067 | alesapin | Unfortunately, this bug was fixed exactly at 19.13.3: https://github.com/ClickHouse/ClickHouse/pull/6748. The only solution for old data -- remove indices and recreate them again.
BTW, it seems like your indices have the wrong definition. Skip indices granularity specified in primary index granularity which by defa... | 2019-09-24T08:26:11 | 2019-09-24T08:36:24 | {} | MEMBER |
ClickHouse | ClickHouse | 534,520,896 | 6,981 | alexey-milovidov | > If you want include contrib to src - it will be not used for packaging and src size will be ~300M
But it is the only right way, because production version of ClickHouse can be build only with specific versions of specific libraries. | 2019-09-24T11:45:35 | 2019-09-24T11:45:35 | {} | MEMBER |
ClickHouse | ClickHouse | 534,570,411 | 7,065 | amosbird | > Can you explain why can't we link the malloc() override into DSO? What is the problem with linker or dynamic loader, if we do otherwise?
Most likely because it might fail to load in time upon starting. | 2019-09-24T13:56:52 | 2019-09-24T13:56:52 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 534,695,279 | 7,026 | akuzm | ```
SELECT MobilePhoneModel, uniq(UserID) AS u FROM hits_100m_single WHERE MobilePhoneModel != '' GROUP BY MobilePhoneModel ORDER BY u DESC LIMIT 10
┏━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━━━┓
┃ version ┃ count() ┃ min ┃ med ┃ max ┃ q95-q05 ┃
┡━━━━━━━━━━━━━━━━━━━━━━━... | 2019-09-24T18:44:32 | 2019-09-24T18:44:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,702,300 | 7,066 | akuzm | Added a comment https://github.com/ClickHouse/ClickHouse/commit/6e32553e0184e151b23dd109a75acfa28630ca4d | 2019-09-24T19:01:57 | 2019-09-24T19:01:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 533,892,783 | 7,009 | dujijun007 | > @dujijun007 Why do you have `clickhosue-server` hostname instead of `clickhouse-server`?
ohhhhhh, you got me! | 2019-09-22T15:43:43 | 2019-09-22T15:43:43 | {} | NONE |
ClickHouse | ClickHouse | 533,926,317 | 6,969 | alexey-milovidov | 70bc895 | 2019-09-22T22:57:27 | 2019-09-22T22:57:27 | {} | MEMBER |
ClickHouse | ClickHouse | 533,946,047 | 6,999 | amosbird | it's https://github.com/yandex/ClickHouse/pull/7049 btw | 2019-09-23T02:52:58 | 2019-09-23T02:52:58 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 533,992,063 | 7,014 | excitoon | @alexey-milovidov That really was an unfinished code, I just based my branch for auth data on my previous branch to make some code here. | 2019-09-23T07:44:31 | 2019-09-23T07:44:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,182,241 | 6,981 | alexey-milovidov | > git submodule update --init cant run on sources downloaded without git.
We should provide source tarballs that are ready to build.
> implicit switch from internal libraries to external ones is a very bad thing. If you want to build without submodules, then explicitly specify `-DUNBUNDLED=ON`
True. One explic... | 2019-09-23T16:40:41 | 2019-09-23T16:40:41 | {} | MEMBER |
ClickHouse | ClickHouse | 534,572,636 | 7,065 | alexey-milovidov | Maybe it is dependency issue. If all .so that do malloc/free depend on .so that provide overrides, this .so must be loaded beforehand. | 2019-09-24T14:01:39 | 2019-09-24T14:01:39 | {} | MEMBER |
ClickHouse | ClickHouse | 534,686,284 | 7,070 | 4ertus2 | 0467229 is bugfix. It should be merged into 19.15. It's better to backport all this PR. | 2019-09-24T18:24:07 | 2019-09-24T18:24:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 533,919,310 | 7,046 | alexey-milovidov | Most likely the index is analyzed slowly and we have potential for improvement. | 2019-09-22T21:13:43 | 2019-09-22T21:13:43 | {} | MEMBER |
ClickHouse | ClickHouse | 533,968,790 | 7,046 | zhang2014 | Because current index scans are single-thread, they are very efficient on indexes like `minmax`, but slow on row-level indexes like `bloomfilter`.
BTW: the progress information is built after the index scan. | 2019-09-23T06:00:33 | 2019-09-23T06:00:33 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,149,535 | 6,914 | maqroll | @nikitamikhaylov @alexey-milovidov got rid of try/catch & added integration test | 2019-09-23T15:23:47 | 2019-09-23T18:09:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,565,323 | 7,083 | alexey-milovidov | `clickhouse-client` receives values of DateTime in binary (unix timestamp). It can use local time zone (when `use_client_time_zone` is set) or time zone received from server (default) to display it in text form.
But when you use `toString` in your query, data transformation happens on server side and the server send... | 2019-09-24T13:45:23 | 2019-09-24T13:46:04 | {} | MEMBER |
ClickHouse | ClickHouse | 534,683,945 | 7,026 | akuzm | ```
SELECT URL, count() AS c FROM hits_100m_single GROUP BY URL ORDER BY c DESC LIMIT 10
┏━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━┳━━━━━━━━━┓
┃ version ┃ count() ┃ min ┃ med ┃ max ┃ q95-q05 ┃
┡━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━╇━━━━━━━╇━━━━━━━╇━━━━━━━━━┩
│ /home/akuzm/... | 2019-09-24T18:18:25 | 2019-09-24T18:18:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 533,925,062 | 7,047 | alexey-milovidov | Ok.
JFYI, let me explain why it requires 2x reservation.
It's not the reservation for the single merge operation.
It's needed to allow other merges to run. For example, first merge will eat 1/2 of free space. Then we can run second merge and it will be able to eat 1/4 of free space. Third merge will be able to eat... | 2019-09-22T22:37:13 | 2019-09-22T22:37:13 | {} | MEMBER |
ClickHouse | ClickHouse | 533,925,928 | 6,969 | alexey-milovidov | Ok forgot that the word "cumulative" is better than "absolute" :(
Will change after merge. | 2019-09-22T22:50:20 | 2019-09-22T22:50:20 | {} | MEMBER |
ClickHouse | ClickHouse | 533,950,856 | 7,050 | jaykelin | ```
2019.09.23 10:43:24.838852 [ 29197 ] {1b97a818-1991-4aa8-8ca8-69239b5129e9} <Error> HTTPHandler: Code: 47, e.displayText() = DB::Exception: Missing columns: 'nodp_play_cd' 'dp' 'dp_ck_cd' 'ck
' 'dp_nock_play' 'dp_nock_cd' 'dp_ck' 'nodp_play' 'cd' 'play' while processing query: 'SELECT scid, dt, hour, code, rank, ... | 2019-09-23T03:37:57 | 2019-09-23T03:37:57 | {} | NONE |
ClickHouse | ClickHouse | 533,951,766 | 7,050 | alexey-milovidov | Does the issue reproduce on version 19.14?
RPM packages are available here: https://repo.yandex.ru/clickhouse/rpm/stable/x86_64/ | 2019-09-23T03:46:15 | 2019-09-23T03:46:15 | {} | MEMBER |
ClickHouse | ClickHouse | 534,226,161 | 6,912 | alexey-milovidov | The idea is Ok, but it's hard to grasp what this function does from its name.
Maybe a different name will be better... | 2019-09-23T18:30:47 | 2019-09-23T18:30:47 | {} | MEMBER |
ClickHouse | ClickHouse | 534,236,572 | 7,067 | alexey-milovidov | Could you please share your table definition?
(The most relevant parts are INDEX definitions) | 2019-09-23T18:57:50 | 2019-09-23T18:57:50 | {} | MEMBER |
ClickHouse | ClickHouse | 534,319,471 | 7,065 | alexey-milovidov | One build has failed. | 2019-09-23T23:09:05 | 2019-09-23T23:09:05 | {} | MEMBER |
ClickHouse | ClickHouse | 534,123,519 | 6,975 | Akazz | Found multiple other cases of "ClickHouse Functional Stateless Tests (thread)" failures:
https://clickhouse-test-reports.s3.yandex.net/7020/b64bab29ed1a329f6f7414f742c2a3a94f4bcf77/functional_stateless_tests_(thread).html
https://clickhouse-test-reports.s3.yandex.net/0/589a3c97438dd5e0c44552cbe223d88f841c0bb0/functio... | 2019-09-23T14:27:09 | 2019-09-23T14:27:09 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,161,055 | 6,981 | abyss7 | @alexey-milovidov mentioned for reference. | 2019-09-23T15:48:57 | 2019-09-23T15:48:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,169,189 | 6,817 | alesapin | @amosbird Yes, this build still broken. Also I have added test for it https://clickhouse-test-reports.s3.yandex.net/0/6b6714761b2cc85c8e28d0cd4df40c90376e06d3/split_build_smoke_test.html (look at test_run log), but now it's always green, because this check is broken. | 2019-09-23T16:08:27 | 2019-09-23T16:08:27 | {} | MEMBER |
ClickHouse | ClickHouse | 534,316,614 | 7,067 | reflection | For context, we saw the `bad size of marks` error with the 19.13.2.19 build, so we tried updating to the newer 19.14.6.12 build, but still saw the same error.
We tried a few things (and still have the old data if needed to help y'all with debugging):
- Tried recreating indices, but were getting errors so had to ... | 2019-09-23T22:57:14 | 2019-09-23T22:57:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,509,111 | 4,689 | alexey-milovidov | Need to update to the recent version. | 2019-09-24T11:08:13 | 2019-09-24T11:08:13 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 534,074,515 | 6,994 | 4ertus2 | Similar issue #6978 | 2019-09-23T12:20:29 | 2019-09-23T12:20:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,082,888 | 7,059 | excitoon | This is unfinished so far (not tested and possibly won't build at all). | 2019-09-23T12:46:16 | 2019-09-23T12:52:59 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,212,465 | 6,161 | akuzm | For the reference, some related changes:
https://github.com/ClickHouse/ClickHouse/pull/6925
https://github.com/ClickHouse/ClickHouse/pull/7062
https://github.com/ClickHouse/ClickHouse/pull/7066 | 2019-09-23T17:56:06 | 2019-09-23T18:17:46 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,270,053 | 6,835 | filimonov | Should be backported to 19.14. Build is broken otherwise in CentOS. | 2019-09-23T20:26:41 | 2019-09-23T20:26:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 534,454,604 | 7,053 | alesapin | Can you provide any data/queries to reproduce the issue? | 2019-09-24T08:35:26 | 2019-09-24T08:35:26 | {} | MEMBER |
ClickHouse | ClickHouse | 534,549,644 | 6,981 | alexey-milovidov | > Its broken
It needs to be fixed.
> In specific ubuntu 18.04 x86_64, but there are dozens of operating systems and architectures in the world
@abyss7 will add cross-compile for Mac OS, FreeBSD, then Linux Aarch64 and Linux Power. | 2019-09-24T13:07:00 | 2019-09-24T13:07:00 | {} | MEMBER |
ClickHouse | ClickHouse | 533,943,933 | 7,006 | flyfoxCI | in fact that my distribute table works well for a long time,but when I recover from the the crush of zookeeper I found the distribute table seems not work. At last I found that when the table has many records in system.replication_queue it will not works. | 2019-09-23T02:32:59 | 2019-09-23T02:32:59 | {} | NONE |
ClickHouse | ClickHouse | 534,566,582 | 7,007 | nicelulu | Work in progress. | 2019-09-24T13:48:16 | 2019-09-24T13:48:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 526,058,934 | 6,707 | filimonov | I think that it will create more problems than benefits. (See comments in PR).
Did you consider HTTP Keep-Alive instead? | 2019-08-29T07:18:01 | 2019-08-29T07:18:01 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.