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 | 344,269,275 | 1,489 | silviucpp | @alexey-milovidov looking to the code in the following function:
```c++
void ReplicatedMergeTreePartCheckThread::start()
{
std::lock_guard<std::mutex> lock(start_stop_mutex);
if (need_stop)
need_stop = false;
else
thread = std::thread([this] { run(); });
}
```
I'm still won... | 2017-11-14T14:06:55 | 2017-11-14T14:06:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,000,264 | 1,503 | alexey-milovidov | BTW, gcc-7 is also Ok. | 2017-11-16T17:38:30 | 2017-11-16T17:38:30 | {} | MEMBER |
ClickHouse | ClickHouse | 344,733,176 | 1,489 | alexey-milovidov | Yes, this looks very suspicious. I will try to figure out. | 2017-11-15T21:24:49 | 2017-11-15T21:24:49 | {} | MEMBER |
ClickHouse | ClickHouse | 345,008,077 | 1,492 | alexey-milovidov | When you type
```
cat file | command > file
```
in your shell,
the shell first opens `file` for writing and truncates it (the `>` redirection) and when `cat` opens the same file for reading, it is already empty.
This has no relation to `clang-format`.
To test clang-format, simply type:
```
clang-format dbm... | 2017-11-16T18:05:52 | 2017-11-16T18:07:08 | {} | MEMBER |
ClickHouse | ClickHouse | 345,501,435 | 1,489 | silviucpp | @alexey-milovidov Yes I agree with you on all 3 dirty places.
I have no idea how we can fix 1 :) and 3 maybe it's better to do it when you knows are no longer other branches with changes in this are to make merges with less conflicts.
I'm happy that you integrated the changes
Have a nice day,
Silviu | 2017-11-19T08:48:57 | 2017-11-19T08:48:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,228,544 | 1,485 | alexey-milovidov | Almost Ok.
But I don't see total report after the patch(the last report after you press Ctrl+C).
I want to compare minimum time:
`0.000% 0.852 sec.` | 2017-11-14T11:22:05 | 2017-11-14T11:22:20 | {} | MEMBER |
ClickHouse | ClickHouse | 344,719,265 | 1,490 | alexey-milovidov | Thanks! | 2017-11-15T20:31:59 | 2017-11-15T20:31:59 | {} | MEMBER |
ClickHouse | ClickHouse | 345,372,639 | 1,489 | silviucpp | @alexey-milovidov ok great !! | 2017-11-17T21:38:55 | 2017-11-17T21:38:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,413,076 | 1,493 | alexey-milovidov | Thank you for the test case!
Fixed in https://github.com/yandex/ClickHouse/commit/14e069f7481664831cab1dce679fa47e21ff901e | 2017-11-18T02:49:05 | 2017-11-18T02:49:05 | {} | MEMBER |
ClickHouse | ClickHouse | 345,426,881 | 1,489 | silviucpp | I merged with your branch and also with master . Was a small conflict because of the resharding removed | 2017-11-18T08:18:45 | 2017-11-18T08:18:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,441,656 | 940 | alexey-milovidov | I think it should be fine! Looking forward for PR. | 2017-11-18T13:15:17 | 2017-11-18T13:15:17 | {} | MEMBER |
ClickHouse | ClickHouse | 344,059,483 | 1,487 | sunsingerus | resolved, RPM-s published here
https://packagecloud.io/altinity/clickhouse
| 2017-11-13T21:12:15 | 2017-11-13T21:12:15 | {} | NONE |
ClickHouse | ClickHouse | 344,257,120 | 1,485 | amosbird | Oh, i didn't notice the last one is a summary :)
## Test without this patch
```
QPS: 1.049, RPS: 183538445.406, MiB/s: 350.072, result RPS: 1.049, result MiB/s: 0.000.
0.000% 0.852 sec.
10.000% 0.879 sec.
20.000% 0.899 sec.
30.000% 0.908 sec.
40.000% 0.917 sec.
50.000% 0.933 sec.
60.000% 0.949 sec.
70.000... | 2017-11-14T13:22:52 | 2017-11-14T13:22:52 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 344,463,427 | 1,485 | alexey-milovidov | ERROR: type should be string, got "https://github.com/yandex/ClickHouse/pull/1495\r\n\r\nNow the code is actually faster than before (+8%).\r\nAlso added functional and performance test." | 2017-11-15T02:10:05 | 2017-11-15T02:10:05 | {} | MEMBER |
ClickHouse | ClickHouse | 344,756,259 | 1,489 | ztlpn | Here is the problem this code is supposed to solve: there are two threads that concurrently start and stop the PartCheckThread - the RestartingThread and the AlterThread. Each of these threads alternates start() and stop() calls, but this is not true in general - we can have two stop() calls and then two start() calls.... | 2017-11-15T22:52:04 | 2017-11-15T22:52:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,139,746 | 1,492 | sundy-li | Got that , thanks | 2017-11-17T04:04:16 | 2017-11-17T04:04:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,544,922 | 1,455 | alexey-milovidov | 👍 | 2017-11-19T20:01:18 | 2017-11-19T20:01:18 | {} | MEMBER |
ClickHouse | ClickHouse | 345,614,462 | 1,517 | amosbird | hmm, http://en.cppreference.com/w/cpp/language/eval_order rule 3 invalidates this patch :) | 2017-11-20T07:39:55 | 2017-11-20T07:39:55 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 344,059,659 | 1,457 | sunsingerus | resolved, RPM-s published here
https://packagecloud.io/altinity/clickhouse
| 2017-11-13T21:12:50 | 2017-11-13T21:12:50 | {} | NONE |
ClickHouse | ClickHouse | 344,325,896 | 1,469 | filimonov | Window functions are not supported currently, but there are chances that they will appear in the future, in some cases you can find workarounds. Possible directions: parametric aggregate functions, higher-order functions, subselects, and LIMIT BY statements.
For running total, there is a related function
https://cl... | 2017-11-14T17:00:45 | 2017-11-14T17:01:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,719,406 | 1,499 | alexey-milovidov | Great! | 2017-11-15T20:32:38 | 2017-11-15T20:32:38 | {} | MEMBER |
ClickHouse | ClickHouse | 344,843,157 | 1,445 | adhamelia | guys any clue ??? still waiting for the response | 2017-11-16T07:54:17 | 2017-11-16T07:54:17 | {} | NONE |
ClickHouse | ClickHouse | 344,905,776 | 1,501 | filimonov | And that requirement looks very specific, and it's very easy to solve that problem without modifications of runningDifference (as i showed), so i don't think that adding that possibility to runningDifference is really needed. | 2017-11-16T12:15:07 | 2017-11-16T12:15:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,037,067 | 1,504 | ludv1x | The described case looks suspicious.
Could you send grep of the log for the precommitted part?
You could restart the server to fix it.
How often does it happen? | 2017-11-16T19:31:36 | 2017-11-16T19:31:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,238,318 | 1,489 | alexey-milovidov | Yes. There is a chance I will do today. | 2017-11-17T13:03:08 | 2017-11-17T13:03:08 | {} | MEMBER |
ClickHouse | ClickHouse | 344,027,518 | 1,059 | proller | @jaykelin Can you prove some repeatable test?
Usually it contain queries: create table; inssert; select;
Now we have
https://github.com/yandex/ClickHouse/blob/master/dbms/tests/queries/0_stateless/00506_union_distributed.sql but it always pass, before and after your changes | 2017-11-13T19:19:40 | 2017-11-13T19:19:40 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,316,572 | 1,470 | filimonov | You need something like that:
```
SELECT
productid,
count() as per_product_count,
( SELECT count() from prod_sales ) as total_count,
per_product_count * 100 /total_count as percent_to_total
FROM prod_sales
GROUP BY productid
```
or the same without extra column:
```
SELECT
productid,
... | 2017-11-14T16:33:57 | 2017-11-14T16:36:11 | {
"+1": 6
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,463,949 | 1,485 | amosbird | Great! | 2017-11-15T02:13:16 | 2017-11-15T02:13:16 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 345,042,337 | 1,504 | vavrusa | It was happening ~5 times a minutes (only on two replicas in the cluster).
DETACH and ATTACH of the table seems to have fixed it, it seems like the whole replication was stuck, as there was a lot of messages like this after attaching:
```
2017.11.16 19:06:59.007812 [ 940 ] <Error> r1.requests_1m (StorageReplicated... | 2017-11-16T19:50:17 | 2017-11-16T19:50:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,368,744 | 1,489 | alexey-milovidov | BTW, I am adding modifications in `silviucpp-replicated_merge_tree_thread_pool` branch (you can merge with it if you will add more commits). | 2017-11-17T21:22:34 | 2017-11-17T21:22:34 | {} | MEMBER |
ClickHouse | ClickHouse | 345,373,804 | 1,489 | silviucpp | @alexey-milovidov one thing I noticed during working on this is that the entire code (in this area at least) it's abusing of friend classes and break the encapsulation. Sometimes when you come from outside and try to read a code it's pretty hard to track where a variable is read/changed as time even if it's private it'... | 2017-11-17T21:44:10 | 2017-11-17T21:44:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,530,776 | 1,515 | alexey-milovidov | Perfect!
BTW, why "Submodule poco updated 265 files"? Probably by accident. I can merge and update it back to the last commit.
> There is a config library that has all this merging and loading logic and is generally more human-friendly. We may think of migrating to it sometime later https://github.com/puppetlabs/... | 2017-11-19T16:45:13 | 2017-11-19T16:45:13 | {} | MEMBER |
ClickHouse | ClickHouse | 345,690,572 | 1,510 | goranc | After upgrade server to 1.1.54310 version we encountered problem with merge speed.
The merges are much slower than on previous stable version.
There were no change in our code, batch size or anything else, and database started to refuse inserts.
I suspect that increased size of blocks to 150MB can be related. It i... | 2017-11-20T13:08:51 | 2017-11-20T13:08:51 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 344,023,955 | 1,446 | alexey-milovidov | We should add a test for the case described in https://github.com/yandex/ClickHouse/issues/1059 | 2017-11-13T19:07:23 | 2017-11-13T19:07:23 | {} | MEMBER |
ClickHouse | ClickHouse | 344,718,991 | 1,491 | vavrusa | @alexey-milovidov I removed it in favour of this. I reopened a PR with just this commit: https://github.com/yandex/ClickHouse/pull/1499 | 2017-11-15T20:30:58 | 2017-11-15T20:30:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,933,204 | 1,503 | proller | can you use newer clang ? (3.9+) | 2017-11-16T14:09:28 | 2017-11-16T14:09:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,592,705 | 1,466 | code-of-kpp | Poco version used: 1.7.6 https://github.com/pocoproject/poco/tree/b35faa0c143221428f0f1fd2c91da2270a20f419
No patches applied to any c or cpp files - only to cmake, regarding iodbc support.
I inserted some debug output, but I don't see anything in logs after recompiling.
Here is what I added:
```diff
void
... | 2017-11-20T05:04:19 | 2017-11-20T05:04:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,719,612 | 1,491 | alexey-milovidov | Ok. | 2017-11-15T20:33:25 | 2017-11-15T20:33:25 | {} | MEMBER |
ClickHouse | ClickHouse | 345,054,636 | 1,504 | vavrusa | There isn't anything particular, this is the first occurence of the part id:
```
2017.11.13 16:03:31.789037 [ 34 ] <Error> DB::StorageReplicatedMergeTree::queueTask()::<lambda(DB::StorageReplicatedMergeTree::LogEntryPtr&)>: Code: 49, e.displayText() = DB::Exception: Could not add part 20171112_20171113_129593_12964... | 2017-11-16T20:36:28 | 2017-11-16T20:45:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,611,758 | 1,510 | AlexPosix | filimonov thanks for reply, we do not use bulk inserts. We inserts for about 280000 per minute. In additional we have another error
2017.11.19 20:00:37.549062 [ 132 ] <Error> HTTPHandler: Code: 60, e.displayText() = DB::Exception: Table default.widget_statistics_summ doesn't exist., e.what() = DB::E
xception, Stack ... | 2017-11-20T07:24:38 | 2017-11-20T07:44:17 | {} | NONE |
ClickHouse | ClickHouse | 343,761,209 | 1,360 | slw | ```
# gdb /NFS/slw/wrk/ClickHouse/dbms/src/Server/clickhouse 48966
GNU gdb 6.1.1 [FreeBSD]
Copyright 2004 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to... | 2017-11-12T19:35:15 | 2017-11-13T14:14:55 | {} | NONE |
ClickHouse | ClickHouse | 344,529,880 | 1,464 | AlexMAS | On Linux the same behaviour. Can anyone help me? May be I need run the container with some privileges? | 2017-11-15T09:08:08 | 2017-11-15T09:08:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,557,085 | 1,491 | proller | maybe redefine thread_local for apple ? | 2017-11-15T10:53:38 | 2017-11-15T10:53:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,716,905 | 1,491 | alexey-milovidov | > it's less intrusive cbf1cff
That's Ok.
BTW I cannot find your commit.
```
milovidov@milovidov-Pro-P30:~/work/ClickHouse$ git fetch --all
Fetching origin
milovidov@milovidov-Pro-P30:~/work/ClickHouse$ git show cbf1cff5bf8c43865364b8981a20a629416b82fc
fatal: bad object cbf1cff5bf8c43865364b8981a20a629416b82f... | 2017-11-15T20:23:08 | 2017-11-15T20:23:19 | {} | MEMBER |
ClickHouse | ClickHouse | 344,734,480 | 1,489 | alexey-milovidov | It looks correct.
`need_stop` set to true is "request to stop".
If requested to stop, we must wait for thread to finish.
When thread is finished, then request to stop is fulfilled and `need_stop` is set to false.
In method `start`, if the "request to stop" is active, we don't start thread and thus request to st... | 2017-11-15T21:29:27 | 2017-11-15T21:29:27 | {} | MEMBER |
ClickHouse | ClickHouse | 344,738,686 | 1,489 | silviucpp | Hello @alexey-milovidov ,
So my notice is that start and end are protected by mutex so concurrent access it's out of discussion.
In start method need_stop can be true only in case stop was called two times in row.
First time the thread being still alive thread.joinable() branch is executed and second time is no... | 2017-11-15T21:45:01 | 2017-11-15T21:45:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,067,240 | 1,489 | silviucpp | @alexey-milovidov @ztlpn there are any tests I can run to check the replication functionality ? I did manual tests including restarting zookeper, replicas and data seems consistent on this branch that's using a pool of threads.
I'll write some tests to see how behaves with 100k tables that are writing in the same ti... | 2017-11-16T21:24:44 | 2017-11-16T21:24:44 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,546,096 | 1,511 | alexey-milovidov | Thanks. Now much better! | 2017-11-19T20:18:03 | 2017-11-19T20:18:03 | {} | MEMBER |
ClickHouse | ClickHouse | 344,224,304 | 1,488 | flowbehappy | 👍 | 2017-11-14T11:05:11 | 2017-11-14T11:05:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,907,484 | 1,501 | michep | Nice solution, thanks. | 2017-11-16T12:23:03 | 2017-11-16T12:23:03 | {} | NONE |
ClickHouse | ClickHouse | 344,957,758 | 1,503 | silviucpp | Hello,
using clang works fine
| 2017-11-16T15:28:48 | 2017-11-16T15:28:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,011,342 | 1,492 | alexey-milovidov | If you are feeling fine to use the most fresh clang version, I would recommend to take a look to the scripts `git-clang-format`, `clang-format-diff`, etc. inside `clang/tools/clang-format` directory:
```
git clone git@github.com:llvm-project/llvm-project-20170507.git
cd llvm-project-20170507
mkdir build && cd bui... | 2017-11-16T18:17:41 | 2017-11-16T18:17:41 | {} | MEMBER |
ClickHouse | ClickHouse | 345,050,483 | 1,504 | ludv1x | > Could you send grep of the log for the precommitted part?
I meant that `grep -e 20171113_20171113_129611_129616_1 -e 20171109_20171114_128172_152787_19 /var/log/clickhouse-server/clickhouse-server.log*` would be helpful to investigate the problem.
I suppose it might be because of two different parallel ZooKeepe... | 2017-11-16T20:20:31 | 2017-11-16T20:36:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,218,322 | 75 | filimonov | +1 to fix | 2017-11-17T11:23:28 | 2017-11-17T11:23:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,324,777 | 64 | harlinb | Potentially solved by #1482 | 2017-11-17T18:30:04 | 2017-11-17T18:30:04 | {} | NONE |
ClickHouse | ClickHouse | 345,538,276 | 1,515 | lemmsh | > BTW, why "Submodule poco updated 265 files"
no idea... I changed nothing there.
HOCON is more convenient than XML and definitely more clear than YAML. We use it as the default config format in our newest systems. It came from Scala world and is probably less known for this reason. Anyway it's just kind of sugar... | 2017-11-19T18:26:05 | 2017-11-19T18:26:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,401,700 | 1,491 | vavrusa | @alexey-milovidov I'm not quite sure yet what's the issue. I couldn't reproduce it with a reduced test case (it build and links), with clickhouse it builds, but doesn't link (the variable symbol is missing). You can use my workaround for the time being, it's less intrusive https://github.com/yandex/ClickHouse/commit/cb... | 2017-11-14T21:18:21 | 2017-11-14T21:18:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,587,736 | 1,464 | tacyuuhon | I also had the same problem. And SELinux is disabled.
But it does work fine on macOS[10.13.1 (17B48)]
Linux
```
$ lsb_release -a
LSB Version: :core-4.1-amd64:core-4.1-noarch
Distributor ID: CentOS
Description: CentOS Linux release 7.3.1611 (Core)
Release: 7.3.1611
Codename: Core
```
... | 2017-11-15T13:10:40 | 2017-11-15T13:37:46 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 344,688,297 | 1,497 | aSealBack | @alexey-milovidov please take a look | 2017-11-15T18:42:26 | 2017-11-15T18:42:26 | {} | NONE |
ClickHouse | ClickHouse | 345,285,761 | 1,510 | vmanyushin | Hello,
have same error after upgrade click house to 54310 build. After reading changelog add this settings to config.xml
` <merge_tree>
<max_bytes_to_merge_at_max_space_in_pool>107374182400</max_bytes_to_merge_at_max_space_in_pool>
</merge_tree>
`
But it doesn't resolve error. Programmers say... | 2017-11-17T16:07:11 | 2017-11-17T16:09:29 | {} | NONE |
ClickHouse | ClickHouse | 345,372,336 | 1,474 | bashkarev | 1.1.54310:
```
[ 8%] Building C object contrib/librdkafka/src/CMakeFiles/rdkafka.dir/crc32c.c.o
[ 8%] Building C object contrib/librdkafka/src/CMakeFiles/rdkafka.dir/rdaddr.c.o
[ 8%] Building CXX object contrib/libtcmalloc/CMakeFiles/tcmalloc_minimal_internal.dir/src/base/atomicops-internals-x86.cc.o
[ 8%] Bui... | 2017-11-17T21:37:33 | 2017-11-17T21:37:33 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 343,881,680 | 1,485 | alexey-milovidov | We should try to test for performance regression. | 2017-11-13T10:46:58 | 2017-11-13T10:46:58 | {} | MEMBER |
ClickHouse | ClickHouse | 344,346,344 | 1,493 | filimonov | Probably caused by that line:
https://github.com/yandex/ClickHouse/blob/fb2d2d36e0ba5b616e81d38061c91eed6184ddfa/dbms/src/Functions/FunctionsStringArray.h#L308 | 2017-11-14T18:07:53 | 2017-11-14T18:07:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,993,231 | 1,500 | proller | dig in progress..
https://github.com/yandex/ClickHouse/blob/master/dbms/src/Interpreters/InterpreterCreateQuery.cpp#L491
https://github.com/yandex/ClickHouse/blob/master/dbms/src/Interpreters/InterpreterSelectQuery.cpp#L297 | 2017-11-16T17:14:35 | 2017-11-16T17:18:02 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,250,234 | 1,504 | ludv1x | Today the problem was reproduced on our production server also.
So, we have all required information now. | 2017-11-17T13:57:06 | 2017-11-17T13:57:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,347,889 | 1,489 | alexey-milovidov | Started doing review... | 2017-11-17T19:47:48 | 2017-11-17T19:47:48 | {} | MEMBER |
ClickHouse | ClickHouse | 345,440,642 | 940 | lemmsh | What if instead of
Poco::AutoPtr<Poco::Util::XMLConfiguration> config = new Poco::Util::XMLConfiguration(config_path); in ExternalLoader.cpp:280 we reuse ConfigProcessor.cpp:375 and so on? I'll try to make a PR a bit later | 2017-11-18T12:56:48 | 2017-11-18T12:56:48 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,489,449 | 1,489 | alexey-milovidov | Functional tests Ok. | 2017-11-19T03:37:07 | 2017-11-19T03:37:07 | {} | MEMBER |
ClickHouse | ClickHouse | 344,096,215 | 1,488 | vavrusa | It's broken since 443088ce3134b1c61e9bc746ed9306bdf0f4c067 (the `extern thread_local` isn't properly supported in the current apple clang). cc @proller | 2017-11-13T23:39:34 | 2017-11-13T23:39:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,145,039 | 1,404 | jaykelin | Merge this pull request, but still unable to start.
then manually delete all `tmp/` in Distributed table dir, is work. | 2017-11-14T04:43:53 | 2017-11-14T04:43:53 | {} | NONE |
ClickHouse | ClickHouse | 344,399,628 | 1,491 | alexey-milovidov | This proposal forces to write `THREAD_LOCAL` instead of `thread_local` - unacceptable.
If LTO is the issue, better to disable LTO on Mac.
... or we can easily switch back to `__thread`. | 2017-11-14T21:10:42 | 2017-11-14T21:10:42 | {} | MEMBER |
ClickHouse | ClickHouse | 344,579,309 | 1,449 | proller | for example freebsd compile: warning on every .cpp:
```
--- dbms/src/Server/CMakeFiles/clickhouse-server.dir/ConfigReloader.cpp.o ---
In file included from /root/ClickHouse/dbms/src/Server/ConfigReloader.cpp:8:
In file included from /root/ClickHouse/dbms/src/Interpreters/Context.h:12:
In file included from /root/C... | 2017-11-15T12:34:54 | 2017-11-15T12:34:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,048,093 | 1,503 | silviucpp | @alexey-milovidov I just updated to last master and last xcode but only clang works.
```
cmake .. -DCMAKE_CXX_COMPILER=`which g++-7` -DCMAKE_C_COMPILER=`which gcc-7` -DCMAKE_CXX_FLAGS="-I/usr/local/include" -DCMAKE_BUILD_TYPE=Debug
```
```
-- The C compiler identification is GNU 7.2.0
-- The CXX compiler ident... | 2017-11-16T20:11:13 | 2017-11-16T20:11:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,099,379 | 1,489 | alexey-milovidov | There are:
Functional tests. Run `dbms/tests/clickhouse-test` with `zookeeper` argument. It will select tests that use Replicated tables. Functional tests are basic - they just perform simple checks for correctness of some isolated functionality. These tests are run on already running server on your local machine.
... | 2017-11-16T23:43:39 | 2017-11-16T23:44:18 | {} | MEMBER |
ClickHouse | ClickHouse | 345,489,440 | 1,489 | alexey-milovidov | Mostly Ok.
There are still some dirty places:
- dependency in ZooKeeper to BackgroundSchedulePool;
- context of MergeSelectingThread in StorageReplicatedMergeTree.cpp (very dirty);
- naming of various Thread classes (need to be renamed to Task). | 2017-11-19T03:36:54 | 2017-11-19T04:36:28 | {} | MEMBER |
ClickHouse | ClickHouse | 344,463,005 | 1,485 | alexey-milovidov | I will totally rewrite this code in separate PR.
But I like to see you in contributors, so I will merge this pull request :) | 2017-11-15T02:07:28 | 2017-11-15T02:07:28 | {} | MEMBER |
ClickHouse | ClickHouse | 344,700,425 | 1,497 | alexey-milovidov | > Places where we intentionally (judging by comments) take the address of the "next after last" character of a string to write '\0' at this address:
./dbms/src/Functions/FunctionsCoding.h:1725: char * pos = &res[0];
./dbms/src/Functions/FunctionsCoding.h:1895: char * pos = &res[0];
./dbms/src/Functions/FunctionsCodi... | 2017-11-15T19:24:11 | 2017-11-15T19:24:11 | {} | MEMBER |
ClickHouse | ClickHouse | 344,721,946 | 1,449 | alexey-milovidov | Let's add `pragma` around `#include <double-conversion...>` | 2017-11-15T20:42:09 | 2017-11-15T20:42:09 | {} | MEMBER |
ClickHouse | ClickHouse | 344,915,723 | 1,489 | silviucpp | @ztlpn thanks ! make sense now | 2017-11-16T13:00:29 | 2017-11-16T13:00:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,934,407 | 1,503 | silviucpp | Build instructions changed ?
I'm using:
cmake .. -DCMAKE_CXX_COMPILER=`which g++-6` -DCMAKE_C_COMPILER=`which gcc-6` -DCMAKE_CXX_FLAGS="-I/usr/local/include" -DCMAKE_BUILD_TYPE=Release
make -j 16 clickhouse
Silviu | 2017-11-16T14:13:56 | 2017-11-16T14:13:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,004,924 | 1,445 | alexey-milovidov | There could be several possible reasons (listed in exception message).
ClickHouse should download correct part from replica and solve the issue automatically.
If you are using quite old ClickHouse release, it's recommended to update to the latest stable. | 2017-11-16T17:54:58 | 2017-11-16T17:54:58 | {} | MEMBER |
ClickHouse | ClickHouse | 345,014,821 | 58 | hmonfared | Any update on this feature ? | 2017-11-16T18:30:01 | 2017-11-16T18:30:01 | {} | NONE |
ClickHouse | ClickHouse | 344,151,690 | 1,485 | amosbird | Thanks. I did some tests.
test setup
```
System: Host: 142.nobida.cn Kernel: 3.10.0-514.26.2.el7.x86_64 x86_64 bits: 64 Console: tty 0
Distro: CentOS Linux release 7.3.1611 (Core)
Machine: Device: server System: Dell product: PowerEdge R720xd serial: N/A
Mobo: Dell model: 01XT2D v: A... | 2017-11-14T05:35:59 | 2017-11-14T05:35:59 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 344,188,445 | 508 | trepik | I have the exact same problem:
`<Error> DB::StorageReplicatedMergeTree::queueTask()::<lambda(DB::StorageReplicatedMergeTree::LogEntryPtr&)>: Code: 33, e.displayText() = DB::Exception: Cannot read all data, e.what() = DB::Exception, Stack trace:`
with a very similar setup (I have 3 nodes for both clickhouse and zooke... | 2017-11-14T08:57:10 | 2017-11-14T08:57:10 | {} | NONE |
ClickHouse | ClickHouse | 344,720,515 | 1,488 | alexey-milovidov | It seems to be fixed.
BTW, it's easy to install the most fresh clang in Mac OS if you have any troubles:
```
git clone git@github.com:llvm-project/llvm-project-20170507.git
cd llvm-project-20170507
mkdir build && cd build
CC=clang CXX=clang++ cmake -D LLVM_TARGETS_TO_BUILD='X86' -D LLVM_ENABLE_PROJECTS='clang... | 2017-11-15T20:36:53 | 2017-11-15T20:36:53 | {} | MEMBER |
ClickHouse | ClickHouse | 344,904,937 | 1,501 | filimonov | For those type of problems modulo calculation is usually used (formula `(x + max_value) % max_value`)
Your sample:
```
:) select x, runningDifference(x) from
(select x from system.one array join [94, 96, 3, 8, 14] as x);
┌──x─┬─runningDifference(x)─┐
│ 94 │ 0 │
│ 96 │ 2 ... | 2017-11-16T12:11:05 | 2017-11-16T12:11:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,057,913 | 1,506 | ludv1x | @vavrusa | 2017-11-16T20:49:26 | 2017-11-16T20:49:26 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,219,927 | 1,474 | proller | Can you try 1.1.54310 ?
lib poco was updated. | 2017-11-17T11:30:58 | 2017-11-17T11:30:58 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,393,122 | 1,489 | alexey-milovidov | Yes, `StorageReplicatedMergeTree` and related classes have broken encapsulation. It because of `StorageReplicatedMergeTree` had became too large and we started to move some code out of it (`ReplicatedMergeTreeAlterThread`, `ReplicatedMergeTreeRestartingThread`, `ReplicatedMergeTreeQueue`, etc), but with no success of r... | 2017-11-17T23:27:31 | 2017-11-17T23:27:31 | {} | MEMBER |
ClickHouse | ClickHouse | 345,440,077 | 940 | lemmsh | This actually used to work in older versions. At least I was able to specify "includes" in the dictionaries xml and they got replaced with actual passwords from the file mentioned in the "<include_from>" directive.
The problem seems to be that ExternalLoader that is used for dictionaries does not do config preproces... | 2017-11-18T12:46:27 | 2017-11-18T12:49:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,087,503 | 1,488 | vavrusa | I just attempted to build it with the same problem, I'll see if it's fixable with current clang. | 2017-11-13T22:56:33 | 2017-11-13T22:56:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,095,308 | 1,485 | alexey-milovidov | How to test:
1. Get a table with many dates.
2. Run `clickhouse-benchmark <<< "SELECT count() FROM test.hits WHERE NOT ignore(toDate(toString(EventDate)))"`
This will do many roundtrip (Date <-> String) transformations in a loop.
3. Keep the results (example):
```
QPS: 28.321, RPS: 251319059.531, MiB/s: 479.353... | 2017-11-13T23:34:52 | 2017-11-13T23:34:52 | {} | MEMBER |
ClickHouse | ClickHouse | 344,165,989 | 1,059 | lppsuixn | @proller
Example:
```sql
CREATE TABLE test_local.union_all_bug on cluster bi_cluster (etldate Date DEFAULT CAST(0 AS Int64),v1 Int64,v2 Int64 DEFAULT CAST(0 AS Int64)) ENGINE =ReplicatedMergeTree('/clickhouse/tables/test/union_all_bug/{shard}', '{replica}', etldate, (etldate), 8192);
CREATE TABLE test.union_al... | 2017-11-14T07:10:02 | 2017-11-14T07:10:02 | {} | NONE |
ClickHouse | ClickHouse | 344,330,950 | 1,491 | vavrusa | This is probably nicer, I'll remove https://github.com/yandex/ClickHouse/pull/1473/commits/cbf1cff5bf8c43865364b8981a20a629416b82fc from my PR 👍 | 2017-11-14T17:17:04 | 2017-11-14T17:17:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,331,362 | 1,491 | vavrusa | Just to make it clear, the only problem is with `extern thread_local` across compilation units (not sure what exactly triggers it, I'm suspecting LTO or something). | 2017-11-14T17:18:19 | 2017-11-14T17:18:19 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 344,462,613 | 1,485 | alexey-milovidov | Ok.
There is 14% performance difference.
Also there is missing test for the modification. | 2017-11-15T02:04:53 | 2017-11-15T02:04:53 | {} | MEMBER |
ClickHouse | ClickHouse | 345,009,425 | 1,492 | alexey-milovidov | Please note that our sources are formatted a little bit different than `clang-format` will do:
We carefully place newlines for long argument lists, long template parameters, ?: operator, etc.
Clang-format config is almost 100% correspond to our style and you can simply use clang-format for new files added by you. B... | 2017-11-16T18:10:59 | 2017-11-16T18:11:38 | {} | MEMBER |
ClickHouse | ClickHouse | 345,190,384 | 1,489 | silviucpp | @alexey-milovidov do you think you can run the integration tests and not available public tests on this branch ? I already rebased all the changes with last master and you can see now as a single commit.
In the meantime I want to write some load tests with lot of tables writing in the same time to see how behaves
... | 2017-11-17T09:26:36 | 2017-11-17T09:26:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,245,717 | 1,489 | silviucpp | @alexey-milovidov btw all tasks that are taking more than 50ms are logged (in order to monitor your concern with threads locked for long time)
Silviu | 2017-11-17T13:37:30 | 2017-11-17T13:37:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 345,269,881 | 1,510 | filimonov | Do you use bulk inserts? How many separate insert statements do you execute per minute/second?
What is in the log files? | 2017-11-17T15:11:24 | 2017-11-17T15:13:34 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.