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 | 269,592,568 | 252 | Nopius | It looks like clickhouse-client now breaks on importing nullable strings starting with '\\'.
Here is my test case:
```
CREATE TABLE test.test2
(
col1 String
) ENGINE = Memory
CREATE TABLE test.test3
(
col1 Nullable(String)
) ENGINE = Memory
```
Now try to populate it with clickhouse-client and 's... | 2016-12-29T07:33:21 | 2016-12-29T07:40:04 | {} | NONE |
ClickHouse | ClickHouse | 271,246,951 | 305 | ymlair | @alexey-milovidov Command dont work，and database engine is mergeTree; maybe it is VPS problem, i reinstall the OS, now it is ok;
i will provide more information if get the problem again,
Thanks. | 2017-01-09T10:02:32 | 2017-01-09T10:02:32 | {} | NONE |
ClickHouse | ClickHouse | 272,103,855 | 324 | jbfavre | Thanks for the answer @alexey-milovidov
I'm preparing a native Debian package and therefore am trying to understand what is important and what's less important :)
Another question: how can I build & run the test suite ?
It doesn't seem to be built by default but I will need it to ensure Debian package quality :)
... | 2017-01-12T08:30:49 | 2017-01-12T08:30:49 | {} | NONE |
ClickHouse | ClickHouse | 273,539,363 | 360 | proller | 2. next errors fixed - https://github.com/yandex/ClickHouse/commit/f63a3ab1674022342f48b7363b0fa1023f5c2ae3 | 2017-01-18T17:17:33 | 2017-01-18T17:17:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,675,882 | 360 | jacktang | @proller `cmake .. -DUSE_STATIC_LIBRARIES=0` works fine. When I step into `make -j $THREADS` command, it throws out `Resource temporarily unavailable` error since I keep default value of `ulimit`. And I try to run `make -j 1` again, here are the error logs
```
...
Scanning dependencies of target example
[ 46%] Bu... | 2017-01-19T03:56:48 | 2017-01-19T03:56:48 | {} | NONE |
ClickHouse | ClickHouse | 269,491,069 | 290 | brigadier | You are right, the problem is that I have yandex.st (and ajax.googleapis.com too, so don't take it personal) blocked in /etc/hosts on my laptop. Because of that jquery won't load and text is not postprocessed. | 2016-12-28T15:19:37 | 2016-12-28T15:19:37 | {} | NONE |
ClickHouse | ClickHouse | 269,875,488 | 294 | alexey-milovidov | It is not in our near plans, but we are interested. Feel free to contribute!
Surely, we will help. | 2016-12-31T17:51:33 | 2016-12-31T17:51:33 | {} | MEMBER |
ClickHouse | ClickHouse | 271,624,705 | 313 | artpaul | I think it has been fixed in master. | 2017-01-10T16:32:44 | 2017-01-10T16:32:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,082,413 | 228 | elukey | Whenever you have time, there are also [previous comments](https://github.com/yandex/ClickHouse/issues/228#issuecomment-268243199) that in my opinion would need some more discussion (cronjobs and permissions chowns hardcoded in init files). | 2017-01-17T10:09:29 | 2017-01-17T10:09:29 | {} | NONE |
ClickHouse | ClickHouse | 271,802,055 | 313 | hatarist | I can confirm that `WHERE ... IN ...` statement is broken, but in my case the problem seems to be weirder:
```
-- `table` has a MergeTree engine.
CREATE TABLE table_buffer AS table ENGINE = Memory;
-- `table_buffer` has exactly the same column names, order etc., but is `Memory`
-- some data was inserted to `ta... | 2017-01-11T07:56:19 | 2017-01-18T09:40:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 269,431,300 | 292 | alexey-milovidov | Ok. But why possibility to set it in AsynchronousMetrics.cpp doesn't work? How you tested it? | 2016-12-28T06:12:34 | 2016-12-28T06:12:34 | {} | MEMBER |
ClickHouse | ClickHouse | 269,466,783 | 290 | alexey-milovidov | Documentation is preprocessed by JavaScript code (located inline, at end of HTML).
If documentation doesn't look properly for you, check that JS is enabled, and there is no JS related errors. You could look for errors in browser console (in Firefox and Chrome, press F12).
Please tell me if there are any JS errors ... | 2016-12-28T11:49:25 | 2016-12-28T11:49:41 | {} | MEMBER |
ClickHouse | ClickHouse | 269,682,272 | 291 | alexey-milovidov | I cannot guess, why it happens.
By the way, what version of OS do you use?
I will try to reproduce... | 2016-12-29T19:41:18 | 2016-12-29T19:41:18 | {} | MEMBER |
ClickHouse | ClickHouse | 271,370,986 | 300 | alexey-milovidov | I still cannot guess, why this issue happens.
Try to isolate possible dependencies. For example, try to enable/disable some types of queries.
Maybe, you could temporary disable queries from Merge table and leave other queries, and see what happens. | 2017-01-09T18:52:25 | 2017-01-09T18:52:25 | {} | MEMBER |
ClickHouse | ClickHouse | 271,438,165 | 241 | alexey-milovidov | Cancelled? | 2017-01-09T23:14:00 | 2017-01-09T23:14:00 | {} | MEMBER |
ClickHouse | ClickHouse | 271,925,722 | 294 | pavel-odintsov | +1 for this thing | 2017-01-11T16:58:36 | 2017-01-11T16:58:36 | {} | NONE |
ClickHouse | ClickHouse | 271,952,238 | 324 | alexey-milovidov | There is no explicit list. I think, every program with INSTALL CMake directive should work.
```
$ find . -name CMakeLists.txt | grep -v contrib | xargs grep -iP 'install\s*\('
./utils/config-processor/CMakeLists.txt:INSTALL(TARGETS config-processor RUNTIME DESTINATION bin COMPONENT config-processor)
./utils/compr... | 2017-01-11T18:26:31 | 2017-01-11T18:26:31 | {} | MEMBER |
ClickHouse | ClickHouse | 272,959,446 | 344 | jbfavre | Using release script already relies on a working debian directory, right ?
Since `release` script generate `debian/control` & `debian/changelog`, and since the build uses (or should use) debhelper, this should work. | 2017-01-16T20:51:54 | 2017-01-16T20:51:54 | {} | NONE |
ClickHouse | ClickHouse | 273,123,968 | 338 | alexeytokar | Thank you for explanation - it makes sense for us :) | 2017-01-17T12:18:00 | 2017-01-17T12:18:00 | {} | NONE |
ClickHouse | ClickHouse | 273,168,094 | 325 | rekby | It isn't work for users with readonly=1 | 2017-01-17T14:02:39 | 2017-01-17T14:02:39 | {} | NONE |
ClickHouse | ClickHouse | 271,555,447 | 238 | proller | looks like it was fixed here https://github.com/yandex/ClickHouse/commit/ca5c1dc31baca3c4645a6322e03e0224a2cd3258
by calling include (cmake/find_libmysqlclient.cmake) with library find
@chipitsine can you try build now with fresh master ?
| 2017-01-10T11:45:53 | 2017-01-10T11:45:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,149,232 | 313 | artpaul | > the _dummy column isn't mentioned anywhere in documentation.
This is internal constant. You can find it in the same file.
> the error now is more than misleading.
Do you have any example of same error which was not fixed? | 2017-01-12T12:16:55 | 2017-01-12T12:17:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,385,706 | 324 | jbfavre | Thanks @alexey-milovidov :) | 2017-01-13T08:15:18 | 2017-01-13T08:15:18 | {} | NONE |
ClickHouse | ClickHouse | 272,946,212 | 338 | alexey-milovidov | > CH behavior on NUMA design when working set for a query will not fit into memory segment of a CPU
ClickHouse allocates most of memory during query execution. Query execution is done with multiple threads and each thread allocates almost equal amout of memory. It means that memory will spread on all available NUMA-... | 2017-01-16T19:38:14 | 2017-01-16T19:39:12 | {
"+1": 4
} | MEMBER |
ClickHouse | ClickHouse | 273,768,786 | 339 | proller | https://github.com/yandex/ClickHouse/commit/22797da2a634e2d6eecf531b4485278fd8bec32c
fixed? | 2017-01-19T12:51:31 | 2017-01-19T12:51:31 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 269,456,174 | 292 | proller | by placing ->GetNumericProperty("tcmalloc.aggressive_memory_decommit" in variuos places:
in start of main, in tcphandler on every query.
and var is true if setted in AsynchrAsynchronousMetrics.cpp to false.
with this patch anything ok. (var always false) | 2016-12-28T10:12:05 | 2016-12-28T10:12:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 269,703,753 | 291 | vavrusa | I know, I was thinking whether it's possible to inject more include paths into compilation here https://github.com/yandex/ClickHouse/blob/master/dbms/src/Interpreters/Compiler.cpp#L201 but it doesn't take anything from env, so probably not.
It would be nice to be able to debug this by injecting additional flags thro... | 2016-12-29T22:31:25 | 2016-12-29T22:32:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,318,052 | 324 | alexey-milovidov | To run functional tests:
1. Start clickhouse-server. Install clickhouse-client or make it visible in $PATH.
2. Go to dbms/tests directory in sources and run `./clickhouse-test`.
In case of success, it will return zero exit code.
Some tests are run only if ZooKeeper is configured in /etc/clickhouse-server/config... | 2017-01-12T23:34:18 | 2017-01-12T23:34:18 | {} | MEMBER |
ClickHouse | ClickHouse | 272,935,109 | 329 | ludv1x | Yes, ClickHouse makes full scan of all MergeTree's parts and applies aggregate function for the column.
This optimization (if column is PK) is possible now even without need to add metainformation for blocks, since MergeTree stores values in sorted order. So it is enough to read first and last row of the column (or ... | 2017-01-16T18:38:25 | 2017-01-16T18:38:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,957,375 | 339 | jbfavre | Here is my solution: https://github.com/yandex/ClickHouse/pull/344 | 2017-01-16T20:40:29 | 2017-01-16T20:40:29 | {} | NONE |
ClickHouse | ClickHouse | 273,551,941 | 360 | proller | 3. - https://github.com/yandex/ClickHouse/commit/a2e87911ce4d55f412501b6e82bfc9f98f745b2b | 2017-01-18T18:02:14 | 2017-01-18T18:02:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,792,288 | 367 | ludv1x | #366 | 2017-01-19T14:37:36 | 2017-01-19T14:37:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,531,452 | 332 | alexey-milovidov | Nice idea!
We have related setting: `distributed_group_by_no_merge`.
If you set it to 1, then intermediate results from remote servers will not be merged.
Example: https://gist.github.com/alexey-milovidov/b097ac3baa210b7743d046e438bf2831
Also it shows, that settings could be specified in subquery, using SETTING... | 2017-01-13T19:52:11 | 2017-01-13T19:52:11 | {} | MEMBER |
ClickHouse | ClickHouse | 273,173,530 | 325 | ludv1x | This query `http://127.0.0.1:8123/?user=readonly&password=123&query=SELECT+number+FROM+system.number+LIMIT+5` works for me with readonly user.
Profile of `readonly` user:
```
<readonly>
<readonly>1</readonly>
</readonly>
``` | 2017-01-17T14:13:19 | 2017-01-17T14:14:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,544,220 | 320 | ludv1x | It looks like a bug in UNION ALL implementation:
```
CREATE TABLE test ( s String, i Int64) ENGINE = Memory
```
```
SELECT s FROM (SELECT s, sum(i*2) AS i FROM test GROUP BY s UNION ALL SELECT s, sum(i*2) AS i FROM test GROUP BY s)
```
Doesn't work:
```
Code: 47. DB::Exception: Received from localhost:900... | 2017-01-18T17:34:30 | 2017-01-18T17:36:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,755,002 | 23 | halilduygulu | for future readers, google translate can translate them to understandable English, i use for specific classes which i can not find in documentation. | 2017-01-19T11:42:26 | 2017-01-19T11:42:26 | {} | NONE |
ClickHouse | ClickHouse | 273,768,686 | 344 | proller | https://github.com/yandex/ClickHouse/commit/22797da2a634e2d6eecf531b4485278fd8bec32c
okay? | 2017-01-19T12:50:59 | 2017-01-19T12:50:59 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,786,123 | 360 | jacktang | @proller I add some debug statements, it seems that f63a3ab commit only fixes cmake error. | 2017-01-19T14:13:24 | 2017-01-19T14:13:24 | {} | NONE |
ClickHouse | ClickHouse | 270,626,246 | 300 | ghost | > Do these values grow with growing of total memory usage?
_MemoryTracking_ metric could be lower or higher, _MemoryTrackingInBackgroundProcessingPool_ is growing (or lowering?), now it has value -253653632. BTW what **-** symbol mean for this value?
> Also, please provide result of
> SELECT * FROM system.merge... | 2017-01-05T11:44:52 | 2017-01-05T11:49:49 | {} | NONE |
ClickHouse | ClickHouse | 271,390,972 | 252 | alexey-milovidov | Confirmed.
Fixed in master. | 2017-01-09T19:59:10 | 2017-01-09T19:59:10 | {} | MEMBER |
ClickHouse | ClickHouse | 271,397,220 | 291 | alexey-milovidov | It is not possible to inject options at runtime. It could be added rather easily. But I suggest first to copy command line and manually find out, what is missing.
In general, `compile` is quite fragile functionality. It even does not always gives significant performance improvement. | 2017-01-09T20:24:42 | 2017-01-09T20:24:42 | {} | MEMBER |
ClickHouse | ClickHouse | 272,082,388 | 238 | chipitsine | I still see "set (MYSQLCLIENT_LIB ${CMAKE_CURRENT_BINARY_DIR}/libmysqlclient.a)" inside "if" and there's nothing like that in "else"
https://github.com/yandex/ClickHouse/blob/master/libs/libmysqlxx/CMakeLists.txt#L32-Lundefined
so,
```
[ilia@localhost ClickHouse]$ cmake .
-- The C compiler identification i... | 2017-01-12T05:52:11 | 2017-01-12T05:52:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 269,639,767 | 232 | hatarist | `USE_STATIC_LIBRARIES=0` is now (690ab98b9) broken, too. :(
```
[ 89%] Linking CXX executable glue_streams
Undefined symbols for architecture x86_64:
"_iconv", referenced from:
DB::FunctionConvertCharset::convert(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std:... | 2016-12-29T14:44:31 | 2016-12-29T14:46:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 270,595,960 | 228 | alexey-milovidov | Sorry for the delay. We have vacations until 9th January.
But I will try to review build issues on Debian sooner. | 2017-01-05T09:06:01 | 2017-01-05T09:06:01 | {} | MEMBER |
ClickHouse | ClickHouse | 271,377,825 | 275 | alexey-milovidov | Thanks for the report! At least we should check for supported cluster names more early and print meaningful error message. | 2017-01-09T19:10:55 | 2017-01-09T19:10:55 | {} | MEMBER |
ClickHouse | ClickHouse | 272,929,392 | 339 | jbfavre | Yes and no.
Yes if you use `release` script.
Probably no when you use native debian packaging.
I will test it, but I think there could be an issue with this: if you have your own fork, there are chances that you add upstream repo as a specific remote. In this case, chances are you also have upstream tags.
But, wh... | 2017-01-16T18:08:31 | 2017-01-16T18:08:31 | {} | NONE |
ClickHouse | ClickHouse | 273,269,151 | 351 | proller | instead of
`create_revision.sh --use_dbms_tcp_protocol_version`
now use:
`cmake .. -DUSE_DBMS_TCP_PROTOCOL_VERSION=1` | 2017-01-17T19:16:15 | 2017-01-17T19:16:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 271,558,079 | 238 | chipitsine | I'll test soon, stay tuned | 2017-01-10T12:00:04 | 2017-01-10T12:00:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,121,957 | 238 | chipitsine | ```
[root@localhost ilia]# ls /usr/lib64/mysql/libmysqlclient*
/usr/lib64/mysql/libmysqlclient.a /usr/lib64/mysql/libmysqlclient.so.18
/usr/lib64/mysql/libmysqlclient_r.so /usr/lib64/mysql/libmysqlclient.so.18.0.0
/usr/lib64/mysql/libmysqlclient.so
[root@localhost ilia]#
```
```
[ilia@localhost ~]$ rpm ... | 2017-01-12T09:59:09 | 2017-01-12T09:59:09 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 269,109,290 | 246 | navdeep710 | Issue got fixed with max_threads setting ... We are running it on VMs
On Dec 20, 2016 9:48 PM, "alexey-milovidov" <notifications@github.com>
wrote:
> Please check the result of query:
>
> SELECT * FROM system.settings WHERE name = 'max_threads'
>
> The query should not return 0. If it returns 0, it indicates a bug.
>... | 2016-12-25T04:01:10 | 2016-12-25T04:01:10 | {} | NONE |
ClickHouse | ClickHouse | 271,225,431 | 200 | artpaul | Спасибо, что раскопали баг. В master исправил. | 2017-01-09T07:44:54 | 2017-01-09T07:44:54 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 271,373,644 | 296 | alexey-milovidov | It means, that server could not connect to specified address in timeout.
Default timeout is quite low: 50 ms (for each of 3 connection tries to each of replica).
If you have geographically distributed cluster, you need to raise the timeout.
It is controlled with `connect_timeout_with_failover_ms` setting.
This ... | 2017-01-09T18:58:23 | 2017-01-09T18:58:23 | {} | MEMBER |
ClickHouse | ClickHouse | 271,857,298 | 318 | alexey-milovidov | Still does not work:
```
-- Performing Test HAVE_GETADDRINFO_A
-- Performing Test HAVE_GETADDRINFO_A - Failed
``` | 2017-01-11T12:33:35 | 2017-01-11T12:33:35 | {} | MEMBER |
ClickHouse | ClickHouse | 272,119,817 | 238 | proller | mysqlclient finding here:
https://github.com/yandex/ClickHouse/blob/master/cmake/find_libmysqlclient.cmake#L6
Did you install libmysqlclient-dev ?
What OS you use ?
can you add line to end of master/cmake/find_libmysqlclient.cmake:
```
message (STATUS "mysqlclient found: MYSQLCLIENT_LIB=${MYSQLCLIENT_LIB}... | 2017-01-12T09:49:42 | 2017-01-12T09:49:42 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 269,684,891 | 291 | vavrusa | I'm not sure how the `-isystem` should work with multiple declarations, is there any way to inject CFLAGS to test whether `-I /usr/...` would help? Everything comes from clang bundled with CH so OS shouldn't matter, but it's Debian 8. | 2016-12-29T19:59:03 | 2016-12-29T19:59:03 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 270,116,374 | 228 | elukey | Gentle ping for the build :) | 2017-01-03T13:42:08 | 2017-01-03T13:42:08 | {} | NONE |
ClickHouse | ClickHouse | 271,403,492 | 253 | alexey-milovidov | We are not going to have native support for NUMERIC/DECIMAL data type in near months.
It is still possible to be implemented in more distant future.
Now we are using UInt64 data type with values, multiplied by fixed amout (like 1 000 000).
For example, we could store price in millionth parts of currency unit. | 2017-01-09T20:50:18 | 2017-01-09T20:50:18 | {} | MEMBER |
ClickHouse | ClickHouse | 271,672,418 | 316 | proller | to ignore depends:
```debuild -d``` | 2017-01-10T19:26:24 | 2017-01-10T19:26:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 271,675,747 | 316 | alexey-milovidov | We should add a flag to `release` script. | 2017-01-10T19:38:34 | 2017-01-10T19:38:34 | {} | MEMBER |
ClickHouse | ClickHouse | 271,962,839 | 323 | ludv1x | More correct performance results obtained using clickhouse-benchmark (0% percentile time):
```
u64 = i64 : 0.102 0.186
u64 >= i64 : 0.106 0.172
i64 > -1 : 0.061 0.062
i64 = 0 : 0.060 0.060
u64 != 0 : 0.060 0.060
i64 = f64 : 0.159 0.134
i64 < f64 : 0.148 0.543
f64 >= 0 : 0.062 0.061
``` | 2017-01-11T19:04:37 | 2017-01-11T19:05:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,938,501 | 228 | elukey | @proller the "-no-pie" flag seems to have fixed the fPIC issue, I have just built the latest master commit and everything went fine. One little thing left: if I do use ldd to check runtime dependencies I can see the following error on Debian Jessie:
```
ldd ./usr/bin/clickhouse-server
./usr/bin/clickhouse-server: ... | 2017-01-16T18:56:04 | 2017-01-16T19:01:58 | {} | NONE |
ClickHouse | ClickHouse | 272,968,844 | 228 | proller | Did you try build without GLIBC_COMPATIBILITY ?
GLIBC_COMPATIBILITY needed if you will run pachage on systems older than jessie
| 2017-01-16T21:43:34 | 2017-01-16T21:43:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,046,081 | 228 | elukey | @proller yes I tried but same issue, GLIBC_2.23 required. I suspect that this is due to the fact that I am building on Debian Sid (to satisfy all the dependencies requested). On which platform do you build the debs that you release? What is the correct procedure?
EDIT: I had a chat with a colleague and this build is... | 2017-01-17T07:59:37 | 2017-01-17T10:06:12 | {} | NONE |
ClickHouse | ClickHouse | 273,536,654 | 360 | ludv1x | @excitoon @proller | 2017-01-18T17:07:56 | 2017-01-18T17:07:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,537,100 | 360 | proller | 1. this os have no libiconv.a : use cmake .. -DUSE_STATIC_LIBRARIES=0
| 2017-01-18T17:09:27 | 2017-01-18T17:09:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 270,601,021 | 300 | ghost | If I understood you right these metrics available from system.metrics table:
```
│ Merge │ 0
│ MemoryTracking │ 21581646
│ MemoryTrackingInBackgroundProcessingPool │ -23638864
```
| 2017-01-05T09:31:51 | 2017-01-05T09:31:51 | {} | NONE |
ClickHouse | ClickHouse | 272,022,676 | 228 | elukey | @proller: in my use case USE_STATIC_LIBRARIES=0 was already working, and I was happy to have shared libraries rather than everything on one binary. The main issue with this setting is that to satisfy dependencies like gcc-6 you'd need to build in Sid for example, precluding the possibility to use the deb created in Jes... | 2017-01-11T22:57:17 | 2017-01-11T22:57:17 | {} | NONE |
ClickHouse | ClickHouse | 272,149,999 | 238 | proller | https://github.com/yandex/ClickHouse/commit/4bfa69765650afcc3345ef1600da63e50918b1c1
Please update and try again | 2017-01-12T12:21:27 | 2017-01-12T12:21:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,334,415 | 313 | bozerkins | i was a release behind. it seems to be working fine now.
sorry for disturbing | 2017-01-13T01:16:53 | 2017-01-13T01:16:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,957,391 | 332 | alexey-milovidov | I remember some issues with building query plan involving ORDER BY or LIMIT.
(like ordered results don't get merged or you get more data than limit)
That is understandable and consistent with behaviour of this option, but may be surprising. | 2017-01-16T20:40:34 | 2017-01-16T20:40:34 | {} | MEMBER |
ClickHouse | ClickHouse | 273,751,350 | 360 | proller | Did you update to current master?
this error-warning should be fixed here https://github.com/yandex/ClickHouse/commit/f63a3ab1674022342f48b7363b0fa1023f5c2ae3 | 2017-01-19T11:25:57 | 2017-01-19T11:25:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,763,809 | 365 | stofel | >Can you try manually follow commands from this script?
Script reinstall many packages and make my devel server complete unworkable.
Now ill try to install gcc6, its not so easy.
| 2017-01-19T12:24:55 | 2017-01-19T12:30:29 | {} | NONE |
ClickHouse | ClickHouse | 269,456,883 | 292 | alexey-milovidov | Ok. | 2016-12-28T10:18:17 | 2016-12-28T10:18:17 | {} | MEMBER |
ClickHouse | ClickHouse | 271,186,929 | 305 | alexey-milovidov | What happens, if you run clickhouse-server from console:
```
sudo service clickhouse-server stop
sudo -u metrika /usr/bin/clickhouse-server --config /etc/clickhouse-server/config.xml
```
What tables are located in `sosa` database? (I am interested in their ENGINE). | 2017-01-08T23:03:17 | 2017-01-08T23:03:17 | {} | MEMBER |
ClickHouse | ClickHouse | 271,624,327 | 314 | artpaul | #313 | 2017-01-10T16:31:29 | 2017-01-10T16:31:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,914,504 | 339 | proller | fixed? e634bc05a9d4e4f07c4225d37be76818a4fa8fa0 | 2017-01-16T16:58:22 | 2017-01-16T16:58:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,934,530 | 332 | re-action | Yeap, you're right, separate function will be more straight forward and understandable approach. However it looks like distributed_group_by_no_merge setting do it's job. Is there any limitations on it's usage? | 2017-01-16T18:35:30 | 2017-01-16T18:35:30 | {} | NONE |
ClickHouse | ClickHouse | 273,123,994 | 228 | proller | > I suspect that this is due to the fact that I am building on Debian Sid (to satisfy all the dependencies requested).
You build on sid and then run on jessie? its bad.
Better find way to install gcc5 or gcc6 on jessie ( https://www.google.ru/webhp?#q=debian%20jessie%20gcc%205 )
>On which platform do you build ... | 2017-01-17T12:18:07 | 2017-01-17T12:18:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,484,522 | 344 | proller | will be fixed here: https://github.com/yandex/ClickHouse/pull/351 | 2017-01-18T14:10:58 | 2017-01-18T14:10:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 269,837,678 | 298 | ludv1x | Updated | 2016-12-31T00:11:58 | 2016-12-31T00:11:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,179,379 | 325 | ludv1x | Anyway special 401 HTTP code for auth fail cases is better than unclear 500 Internal Server Error. | 2017-01-17T14:26:27 | 2017-01-17T17:15:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,791,097 | 354 | ludv1x | Sent detailed answer on your email.
It seems that in this example order of fields inside `SELECT (..fields..)` is incorrect in your query.
At least, in a similar example from your letter it was so (table `session_all_bug`).
| 2017-01-19T14:33:01 | 2017-01-19T14:33:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 271,365,499 | 300 | alexey-milovidov | > BTW what - symbol mean for this value?
It means, that memory was freed in context of background job (like eviction of old data from cache), but was allocated in other context. Usually, this is fine.
I had hypothesis, that memory usage is due to merges, but your data shows that this is not true.
| 2017-01-09T18:33:43 | 2017-01-09T18:33:43 | {} | MEMBER |
ClickHouse | ClickHouse | 271,524,705 | 241 | artpaul | Yes, there is some another cons with this proposal. So it's better to keep as is now. | 2017-01-10T09:17:35 | 2017-01-10T09:17:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 271,918,466 | 300 | ghost | Unfortunately we can't disable queries to Merge table since the bug happened on our prod environment.
Instead we had tried temporary switch off all write activity to ClickHouse and got the similar picture for RAM usage:
. As stated above, I'll follow what you mentioned trying to build and use gc... | 2017-01-17T12:47:30 | 2017-01-17T12:47:47 | {} | NONE |
ClickHouse | ClickHouse | 273,163,175 | 325 | ludv1x | > It need for allow direct access to clickhouse from browsers, they doesn't send login,password for url queries in form http://user:password@server
You could use `http://localhost:8123/?user=default&password=password&query=SELECT+123` instead of `http://user:password@server`. | 2017-01-17T13:47:22 | 2017-01-17T13:47:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,750,864 | 365 | proller | Checked now - anything works for me.
Can you try manually follow commands from this script?
Also here almost ready freebsd port:
https://github.com/proller/clickhouse-freebsd/tree/master/ports/databases/clickhouse
usage:
```
git clone https://github.com/proller/clickhouse-freebsd.git
cd clickhouse-freebsd/... | 2017-01-19T11:23:48 | 2017-01-19T11:23:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,762,614 | 364 | ludv1x | Ok, LGTM :+1: | 2017-01-19T12:18:38 | 2017-01-19T12:18:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,783,921 | 360 | jacktang | yes, I pulled the latest code and rebuilt it
```
$ git show
commit ebf3235282eff9d7af9cb1745761ef0cfe43a7ad
Author: Alexey Zatelepin <ztlpn@yandex-team.ru>
Date: Wed Jan 18 19:04:55 2017 +0300
INTO OUTFILE documentation [#CLICKHOUSE-2133]
``` | 2017-01-19T14:04:26 | 2017-01-19T14:04:26 | {} | NONE |
ClickHouse | ClickHouse | 269,699,907 | 291 | alexey-milovidov | In case of error, temporary .cpp file will reside in its directory, and you could reproduce the error by just copy-pasting and running a command from erorr log:
```
LD_LIBRARY_PATH=/usr/share/clickhouse/bin/ /usr/share/clickhouse/bin/clang -B /usr/share/clickhouse/bin/ -x c++ -std=gnu++1y -O3 -g -Wall -Werror -Wnon... | 2016-12-29T21:57:19 | 2016-12-29T21:58:24 | {} | MEMBER |
ClickHouse | ClickHouse | 269,843,130 | 293 | alexey-milovidov | Ok. But missing test case when LIMIT BY keys are not contiguous in result set.
It is important, because if we assume, that keys are contiguous, we could use light streaming O(1) implementation of LIMIT BY. But we consciously use implementation with O(n) memory to support that case. | 2016-12-31T01:57:20 | 2016-12-31T01:57:20 | {} | MEMBER |
ClickHouse | ClickHouse | 270,596,610 | 300 | alexey-milovidov | Please show graphs of following metrics:
ClickHouse.Metrics.MemoryTracking
ClickHouse.Metrics.MemoryTrackingInBackgroundProcessingPool
ClickHouse.Metrics.Merge | 2017-01-05T09:09:28 | 2017-01-05T09:09:28 | {} | MEMBER |
ClickHouse | ClickHouse | 270,619,848 | 300 | alexey-milovidov | Do these values grow with growing of total memory usage?
Also, please provide result of
`SELECT * FROM system.merges`
query at some time after server startup.
What kind of tables do you use?
If you also provide structure of corresponding tables, it will help. | 2017-01-05T11:06:42 | 2017-01-05T11:06:42 | {} | MEMBER |
ClickHouse | ClickHouse | 271,678,554 | 228 | proller | some fixes merged https://github.com/yandex/ClickHouse/pull/316/
now in debian jessie works: ```debuild -e USE_STATIC_LIBRARIES=0``` | 2017-01-10T19:48:51 | 2017-01-10T19:48:51 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,959,008 | 344 | proller | Now it depends on building from debian/rules.
We decide commit current version with sources, i will prepare patch in next few days.
| 2017-01-16T20:49:19 | 2017-01-16T20:49:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 272,960,786 | 344 | proller | release script - yes, but libs/libcommon can be compiled in any os without /release script | 2017-01-16T20:59:16 | 2017-01-16T20:59:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 273,552,187 | 346 | ludv1x | Thank you for you note.
NULL functionality is unstable and not finalized yet (especially for JOIN clauses).
Finalized functionality will include proper JOIN implementation. | 2017-01-18T18:03:05 | 2017-01-18T18:03:15 | {
"+1": 3
} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.