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 | 518,752,950 | 6,325 | kreuzerkrieg | @alexey-milovidov what are CMake flags to use for unbudled build like in the CI, I will check unbundled and the GCC 9 problems
In addition, how to run clang based builds? is there any flags to apply or just set compiler as clang for the build?
| 2019-08-06T16:51:15 | 2019-08-06T16:51:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,911,931 | 6,369 | den-crane | TTL for Columns and Tables -- Determines the lifetime of values.
https://clickhouse.yandex/docs/en/operations/table_engines/mergetree/#table_engine-mergetree-ttl | 2019-08-07T02:16:42 | 2019-08-07T02:16:42 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,918,311 | 6,148 | zhang2014 | has been merged in #6338 @blinkov | 2019-08-07T02:51:10 | 2019-08-07T02:51:10 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,987,360 | 5,710 | filimonov | Currently that masking is applied only on logs and when putting query to query_log and process list, but can easely be extended to more places.
And yes, you need to put password, or some good enough regexp to match it in config file. | 2019-08-07T08:01:54 | 2019-08-07T08:01:54 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,881,506 | 6,248 | den-crane | minimal STR:
```
CREATE TABLE myTable (`id` String)
ENGINE = MergeTree ORDER BY id SETTINGS index_granularity = 8192;
insert into myTable select toString(number) from numbers(5000);
insert into myTable select '439385272';
select id, toFloat64(id) = 439385272.0 as compare_indicator from myTable
where toFloat6... | 2019-08-06T23:33:36 | 2019-08-06T23:33:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,074,727 | 6,371 | 4ertus2 | Trying to reproduce in our CI #6378 | 2019-08-07T12:28:53 | 2019-08-07T12:28:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,536,064 | 6,334 | 4ertus2 | fixed in master | 2019-08-08T14:13:49 | 2019-08-08T14:13:49 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,594,181 | 6,336 | bopohaa | This is possible when partitions are not filled evenly. Because all consumers for one math view read in one stream sequentially | 2019-08-06T09:40:02 | 2019-08-06T09:40:02 | {} | NONE |
ClickHouse | ClickHouse | 518,604,586 | 6,336 | igor-sh8 | Hi.
All partitions are filled evenly and I see that lag increased on all partitions in the topic.
And on the producer side it is random partitioner. | 2019-08-06T10:07:19 | 2019-08-06T10:07:19 | {} | NONE |
ClickHouse | ClickHouse | 518,750,593 | 6,334 | astudnev | Actually just SELECT ... ORDER BY decimal_field does not work for tables more than 100K records, where float64 works with no problem | 2019-08-06T16:45:15 | 2019-08-06T16:45:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,877,258 | 6,248 | den-crane | I failed to reproduce it with select number from numbers only with rand and partially
```
CREATE TABLE myTable (`id` String)
ENGINE = MergeTree ORDER BY id SETTINGS index_granularity = 8192;
insert into myTable select toString(rand()) from numbers(10000000);
insert into myTable select 439385272;
1.
select id... | 2019-08-06T23:13:08 | 2019-08-06T23:14:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,920,152 | 6,338 | zhang2014 | sorry very much, I didn't notice that 😥 | 2019-08-07T03:00:57 | 2019-08-07T03:03:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,040,272 | 6,355 | akuzm | We use submodules, so after checking out a branch, you have to update them:
```
git submodule sync
git submodule update --init --recursive
```
After that, `git status` should be clean, in particular, it should not show lines like `M contrib/hyperscan`, that mean that a submodule is modified. | 2019-08-07T10:32:35 | 2019-08-07T10:32:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,151,588 | 6,377 | MeteHanC | > > So you mean I should not let CH to distribute queries and query the nodes raund-robin with using ha-proxy or ch-proxy. So I will be querying the local tables instead of distributed ones, did I get you right?
>
> Right.
By the way about this one, I won't have replicas only, I will have shards too. So not letti... | 2019-08-07T15:36:34 | 2019-08-07T15:36:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,207,930 | 6,378 | 4ertus2 | There's no alignment in AllocatorWithStackMemory. It leads to crash in attepmt to push_back __int128 into PODArrayWithStackMemory | 2019-08-07T18:05:13 | 2019-08-07T18:05:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,468,541 | 6,397 | akuzm | CH 19.7 is using lz4 1.9.0, and some out-of-bound reads were introduced in this release, so this might be related. CH 19.11 is on the latest lz4 1.9.1, which should have these fixed.
The first stack is the thread that receives SIGSEGV (824). The log message is apparently from another thread (833) that received SIGAB... | 2019-08-08T10:48:25 | 2019-08-08T10:48:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,668,571 | 6,355 | akuzm | Could you try to clone this repository and build it with cmake, as described here https://clickhouse.yandex/docs/en/development/build/ ? The last stable version is at a git tag `v19.11.5.28-stable`. If it still fails, please post the cmake invocation and its output. | 2019-08-06T13:26:23 | 2019-08-06T13:26:23 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,849,097 | 6,357 | den-crane | And CH needs a Warp drive as well. | 2019-08-06T21:22:16 | 2019-08-06T21:22:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,058,308 | 6,377 | MeteHanC | > did you try `set prefer_localhost_replica = 0` ?
I set prefer_localhost_replica to 0 as you suggested but still no luck. Tried setting 'load_balancing' in users.xml to 'random' again, but it did not work as well. | 2019-08-07T11:35:01 | 2019-08-07T11:35:01 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,087,263 | 6,377 | den-crane | it's a bad idea to send all queries through one server. You should use some load balancer in front of CH and query all nodes (round-robin or random). Try ha-proxy or ch-proxy (https://github.com/Vertamedia/chproxy).
It's bad for high availability.
It's bad for performance because you entrance node re-send a query... | 2019-08-07T13:04:40 | 2019-08-07T13:07:37 | {
"+1": 2
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,157,689 | 6,377 | den-crane | How to set load_balancing it depends of your cluster design.
I use load_balancing = nearest_hostname because I have a geo-cluster (replicas are far (in distance) from each other (high latency/low bandwidth).
Shards have similar host-names inside DC.
And nearest_hostname allows to query (preferably) local (in curre... | 2019-08-07T15:51:39 | 2019-08-07T15:52:54 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,379,803 | 6,365 | filimonov | Now test-stand configuration and test code itself if mixed and tightly coupled.
The main idea of that concept - is untight that, and allow to create new configurations fast and easily.
I.e. split test stand configuration from test code.
When you have such a test-stand configuration described in docker-compose.ym... | 2019-08-08T06:09:15 | 2019-08-08T06:09:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,551,533 | 6,383 | qietingfengling | thanks ,I will try it | 2019-08-08T14:49:12 | 2019-08-08T14:49:12 | {} | NONE |
ClickHouse | ClickHouse | 518,620,921 | 6,320 | alexey-milovidov | @vitlibar Let you continue. We need to modify CMakeLists. | 2019-08-06T10:57:59 | 2019-08-06T10:57:59 | {} | MEMBER |
ClickHouse | ClickHouse | 518,750,843 | 6,334 | astudnev | MergeTree is the same | 2019-08-06T16:45:53 | 2019-08-06T16:45:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,751,986 | 6,366 | alesapin | 1. --Need to add alter for storage Kafka.--
2. I'm afraid of race conditions for MergeTree settings:
For `ALTER MODIFY SETTING` we lock only `alterIntentionLock`. But some background tasks (merges, mutations) can read settings values during this `ALTER`. It may lead to stale read for numeric settings (it's ok). But ... | 2019-08-06T16:48:44 | 2019-08-07T13:36:52 | {} | MEMBER |
ClickHouse | ClickHouse | 519,214,222 | 6,378 | alexey-milovidov | Also we can replace `__int128` to more portable `boost::cpp_int`. It may lead to performance degradation, but most likely not. | 2019-08-07T18:23:02 | 2019-08-07T18:23:02 | {} | MEMBER |
ClickHouse | ClickHouse | 519,450,863 | 6,215 | peaksnail | Hi
I find that when payMode.fee equals [0], it will get the empty array
sql like
```
select arrayReduce('sumMap', array([100]), array([0]))
```
return
```
([],[])
```
but I want get the result
```
([100],[0])
``` | 2019-08-08T09:51:40 | 2019-08-08T09:52:01 | {} | NONE |
ClickHouse | ClickHouse | 519,495,017 | 6,397 | akuzm | I've looked at the commits, but since we don't know much about your case, I can't say whether they are relevant. We'll update when they release a new version. If you wish to test these fixes now, it should be possible to compile with their dev version by just checking out the dev branch in contrib/lz4. | 2019-08-08T12:21:31 | 2019-08-08T12:21:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,582,586 | 6,401 | den-crane | Right, I've missed () for the tuple.
And yes segfault should be fixed somehow. Maybe forbiden as wrong syntax at create table level. | 2019-08-08T16:04:55 | 2019-08-08T16:06:37 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,688,509 | 6,348 | Vxider | I have encountered this issue by using `-D ENABLE_POCO_NETSSL=0`. | 2019-08-06T14:16:11 | 2019-08-06T14:16:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,759,792 | 6,336 | bopohaa | We use protobuff as a format and have an insertion rate from one consumer (thread) of more than 100,000 messages per second (the kafka server is located in another city) | 2019-08-06T17:08:58 | 2019-08-06T17:08:58 | {} | NONE |
ClickHouse | ClickHouse | 518,865,521 | 6,325 | alexey-milovidov | After you click at the build check:
https://clickhouse-builds.s3.yandex.net/6325/4044612832715f4761ee2d3ab82e23e03abec6bb/report.html
you can find build logs for every compiler, for example:
https://clickhouse-builds.s3.yandex.net/6325/4044612832715f4761ee2d3ab82e23e03abec6bb/build_log_480503911_1564946647.txt
... | 2019-08-06T22:21:04 | 2019-08-06T22:21:18 | {} | MEMBER |
ClickHouse | ClickHouse | 518,878,794 | 6,248 | den-crane | hhmm. Wow.
OPTIMIZE TABLE myTable FINAL; did the trick
```
select id, toFloat64(id) = 439385272.0 as compare_indicator from myTable
where toFloat64(id) between 439385271.999 and 439385272.001
0 rows in set. Elapsed: 0.002 sec.
Key condition: (toFloat64(column 0) in [439385271.999, +inf)), (toFloat64(column ... | 2019-08-06T23:20:24 | 2019-08-06T23:21:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,028,883 | 6,374 | dimarub2000 | Tests will be added soon. | 2019-08-07T09:57:14 | 2019-08-07T09:57:14 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,107,272 | 5,825 | simPod | @4ertus2 we upgraded to 19.13.1.11 and it's fixed there. Should have been somewhere between that and 19.9.2.4 | 2019-08-07T13:52:29 | 2019-08-07T13:53:03 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,382,437 | 6,335 | astudnev | There are several servers that do not have this cluster in remote_servers. Is it a problem?
Does it means that remote_servers must declare all clusters on all servers even if they do not participate in that clusters? | 2019-08-08T06:19:06 | 2019-08-08T06:19:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,469,460 | 6,393 | akuzm | What does `dmesg` say, is oom killer invoked? | 2019-08-08T10:51:43 | 2019-08-08T10:51:43 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,549,011 | 6,401 | den-crane | Why did you use [ ] -- it's an array not a tuple
try `ENGINE = SummingMergeTree(price, spend)`
and by the way `clicks` won't be summed, you'll get `any(clicks)` -- are you sure?
ENGINE = SummingMergeTree([columns]) -- [ ] - it means columns are optional (documentation agreement). | 2019-08-08T14:43:18 | 2019-08-08T14:54:20 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,700,244 | 6,338 | alexey-milovidov | I have reverted your commit, please make another PR. | 2019-08-06T14:43:05 | 2019-08-06T14:43:05 | {} | MEMBER |
ClickHouse | ClickHouse | 518,726,059 | 6,336 | igor-sh8 | In this case we will have too small load on the one shard (in our logic schema we have only a few topics) and we will get the overhead in Clickhouse shards.
In our situation we would like to increase the load on the one shard.
Is there some way to set a thread per partition or maybe a few threads on a topic ?
| 2019-08-06T15:41:39 | 2019-08-06T15:41:39 | {} | NONE |
ClickHouse | ClickHouse | 519,053,528 | 6,377 | javisantana | did you try `set prefer_localhost_replica = 0` ? | 2019-08-07T11:17:38 | 2019-08-07T11:17:38 | {
"+1": 3
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,100,066 | 6,366 | alesapin | Ok, first iteration without Kafka settings alter. | 2019-08-07T13:36:29 | 2019-08-07T13:36:29 | {} | MEMBER |
ClickHouse | ClickHouse | 519,480,393 | 6,393 | akuzm | It's strange that the log is almost empty, let's set `logger.level = "trace"` so that we have more messages (https://clickhouse.yandex/docs/en/operations/server_settings/settings/#server_settings-logger).
Does it run long enough so that you connect with clickhouse-client? What's in `system.merges`? I've recently see... | 2019-08-08T11:31:24 | 2019-08-08T11:31:24 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,661,939 | 6,359 | sundy-li | Maybe the Block TTL feature for Storage Memory could help you.
If it's really useful, I will make a simple PR.
```
ubuntu-172.21.214.52 :) CREATE TABLE default.testa (`number` UInt64) ENGINE = Memory() settings memory_max_data_lifetime = 3;
CREATE TABLE default.testa
(
`number` UInt64
)
ENGINE = Memo... | 2019-08-06T13:09:16 | 2019-08-06T13:09:16 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,046,941 | 6,341 | simonhou | @den-crane This is the configuration of http port.
```
[root@~]# cat /etc/ck9024/config.xml | grep http
<http_port>10024</http_port>
```
Besides, I can get results using select * from system.events. Are some events removed in 19.6.2? | 2019-08-07T10:55:12 | 2019-08-07T10:55:12 | {} | NONE |
ClickHouse | ClickHouse | 519,077,500 | 6,248 | dimarub2000 | Wanted to mention that cast from `String` to any `"number"` data type is not monotonic. So all the rows will be read anyway. | 2019-08-07T12:36:52 | 2019-08-07T12:48:05 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,099,706 | 6,377 | den-crane | >By the way round-robin is 'in_order' attribute in the settings right?
No. https://clickhouse.yandex/docs/en/operations/settings/settings/#load_balancing-in_order
the same order as they are specified in configuration
<replica>
<host>first</host><port>9000</port>
... | 2019-08-07T13:35:38 | 2019-08-07T13:35:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,530,304 | 6,401 | SaltTan | Does it crash if you remove square brackets?
SummingMergeTree(price, spend) | 2019-08-08T13:59:40 | 2019-08-08T13:59:40 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,757,352 | 6,336 | bopohaa | 1. Change the message format to reduce the cost of deserialization (protobuf, csv, native)
2. Several view on the same topic, one consumer for each view | 2019-08-06T17:02:25 | 2019-08-06T17:02:25 | {} | NONE |
ClickHouse | ClickHouse | 518,798,491 | 6,366 | alex-zaitsev | > 1. Need to add alter for storage Kafka.
We need to be careful with that and probably do it separately. Some settings can be altered but others can not. | 2019-08-06T18:55:04 | 2019-08-06T18:55:04 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,144,882 | 6,377 | MeteHanC | > Yes, this is expected behavior. You can execute [touch cluster_config.xml ] to reset error counters until #5317 is not implemented
I see, so when I restart the clickhouse cluster on the leader node, it resets error counters as well, because it gets fixed with that way too. Thanks a lot for your time man, I appreci... | 2019-08-07T15:20:36 | 2019-08-07T15:20:36 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,195,023 | 6,335 | filimonov | If one server does not participate in several clusters you should not get the error. | 2019-08-07T17:31:11 | 2019-08-07T17:31:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,555,769 | 6,394 | alesapin | We will enable this setting by default. | 2019-08-08T14:58:59 | 2019-08-08T14:58:59 | {} | MEMBER |
ClickHouse | ClickHouse | 518,702,308 | 6,357 | filimonov | Thank you for the feedback, it's very valuable.
There are plans addressing most of listed problems.
Some of plans are long-term and not very concrete, others (like resource pools) we expect to be delivered still in 2019.
Here is the list of short-term plans: https://github.com/yandex/ClickHouse/issues/4785
Mo... | 2019-08-06T14:47:34 | 2019-08-06T14:47:34 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,705,835 | 6,336 | bopohaa | Right, but if you make several CH shards, each of which will be a consumer of a topic with several partitions, then we will get good scalability. | 2019-08-06T14:55:11 | 2019-08-06T14:55:11 | {} | NONE |
ClickHouse | ClickHouse | 518,878,418 | 4,785 | alexey-milovidov | October 2019. | 2019-08-06T23:18:41 | 2019-08-06T23:18:41 | {} | MEMBER |
ClickHouse | ClickHouse | 519,095,673 | 6,377 | MeteHanC | > it's a bad idea to send all queries through one server. You should use some load balancer in front of CH and query all nodes (round-robin or random). Try ha-proxy or ch-proxy (https://github.com/Vertamedia/chproxy).
>
> It's bad for high availability.
>
> It's bad for performance because you entrance node re-se... | 2019-08-07T13:25:08 | 2019-08-07T13:25:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,509,098 | 6,332 | abyss7 | Fixed overall kafka tests: #6354
Thank you for the attempt. | 2019-08-08T13:04:22 | 2019-08-08T13:04:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,531,880 | 6,401 | dgrr | It doesn't work without brackets and following the Clickhouse doc:
`columns - a tuple with the names of columns where values will be summarized`
So I just want to sum the price and spend columns | 2019-08-08T14:03:28 | 2019-08-08T14:03:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,712,667 | 6,308 | splichy | @abyss7 same on version 19.11.5.28 | 2019-08-06T15:10:34 | 2019-08-06T15:10:34 | {} | NONE |
ClickHouse | ClickHouse | 519,068,901 | 6,377 | MeteHanC | Okay, so I've set prefer_localhost_replica to 0 on CLI, but it seems to not saving the setting. I closed the CLI and re-opened it to double check, I saw that 'prefer_localhost_replica' is re-set to 0.
I have solved the issue by adding `<prefer_localhost_replica>0</prefer_localhost_replica>` line to users.xml. I wil... | 2019-08-07T12:10:50 | 2019-08-07T12:10:50 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,071,258 | 6,336 | igor-sh8 | We use Native format in our schema.
And also for now we use a few MatViews per table.
Ideally we just wanted to make the whole architecture more evident.
In case with MatViews, when we will add a few more tables, we will need to create a whole set of MatViews for each table. It is a bit uncomfortable.
But in any... | 2019-08-07T12:17:41 | 2019-08-07T12:17:41 | {} | NONE |
ClickHouse | ClickHouse | 519,314,450 | 6,355 | alexey-milovidov | You can debug this issue.
1. With gdb:
`gdb --args affected command`
In gdb, type `catch throw`.
Then type `r`.
When it will say that an exception was thrown, type `bt`.
2. With strace:
`strace -f affected command`
Look at the latest `mmap` syscall. | 2019-08-08T00:18:39 | 2019-08-08T00:18:39 | {} | MEMBER |
ClickHouse | ClickHouse | 519,471,446 | 6,393 | Chocorean | It looks like it does kill clickhouse-server. There is a short extract from `dmesg` output:
```
[ 5683.514206] Out of memory: Kill process 1887 (clickhouse-serv) score 979 or sacrifice child
[ 5683.514305] Killed process 1887 (clickhouse-serv) total-vm:22897712kB, anon-rss:16065748kB, file-rss:0kB, shmem-rss:0kB
[ ... | 2019-08-08T10:58:52 | 2019-08-08T10:58:52 | {} | NONE |
ClickHouse | ClickHouse | 519,486,461 | 6,397 | thomas955 | Ok we are trying to reproduce as best as we can.
Could you please double check the commits i mentioned above. These commits are currently _not_ in the master branch of LZ4. | 2019-08-08T11:52:49 | 2019-08-08T11:52:49 | {} | NONE |
ClickHouse | ClickHouse | 519,507,098 | 6,330 | abyss7 | Fixed in another way: #6354 | 2019-08-08T12:58:41 | 2019-08-08T12:58:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,862,730 | 6,287 | den-crane | @dkhodorchenko can you show radacct DDL ? partition / orderby section.
I suspect this is a duplicate for https://github.com/yandex/ClickHouse/issues/5131 | 2019-08-06T22:09:22 | 2019-08-06T22:09:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,085,599 | 6,334 | 4ertus2 | Reproduced. No inserts in MergeTree with Decimal key.
```
DROP TABLE IF EXISTS t1;
DROP TABLE IF EXISTS t2;
CREATE TABLE t1 (str String, dec Decimal64(8)) ENGINE = MergeTree ORDER BY str;
CREATE TABLE t2 (str String, dec Decimal64(8)) ENGINE = MergeTree ORDER BY dec;
SELECT now();
INSERT INTO t1 SELECT toS... | 2019-08-07T13:00:18 | 2019-08-07T13:00:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,104,203 | 6,377 | MeteHanC | > > By the way round-robin is 'in_order' attribute in the settings right?
>
> No. https://clickhouse.yandex/docs/en/operations/settings/settings/#load_balancing-in_order
> the same order as they are specified in configuration
>
So you mean I should not let CH to distribute queries and query the nodes raund-robin... | 2019-08-07T13:46:47 | 2019-08-07T13:46:47 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,108,557 | 6,334 | 4ertus2 | #6380 | 2019-08-07T13:55:27 | 2019-08-07T13:55:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,129,263 | 6,377 | den-crane | > So you mean I should not let CH to distribute queries and query the nodes raund-robin with using ha-proxy or ch-proxy. So I will be querying the local tables instead of distributed ones, did I get you right?
Right.
> By the way at the moment I am having another issue, when I restart a CH replica under load... | 2019-08-07T14:44:56 | 2019-08-07T14:44:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,130,903 | 6,383 | den-crane | First of all your table should be partitioned by day. And the answer depends of your partitioning expression PARTITION BY toYYYYMMDD(date) / PARTITION BY date.
alter table xxx drop partition 20190807 / alter table xxx drop partition '2019-08-07'
Also check TTL https://clickhouse.yandex/docs/en/operations/table... | 2019-08-07T14:48:37 | 2019-08-07T14:49:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,622,307 | 6,336 | bopohaa | In your case, when reading in 6 threads, everything is fine, and when reading in one thread, there is a lack. It can be assumed that there is not enough performance for 6 consumers to work on one processor core. See how the load on the processor cores is distributed. | 2019-08-06T11:02:25 | 2019-08-06T11:02:25 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 518,642,730 | 6,341 | simonhou | @filimonov Thanks for your replying. I've tried as what you supported. There is still no matching data returned. The CH version is 19.6.2.1 . I download the rpm package form https://packagecloud.io/Altinity/clickhouse. | 2019-08-06T12:12:14 | 2019-08-06T12:12:14 | {} | NONE |
ClickHouse | ClickHouse | 519,074,306 | 6,368 | millin | I'm not sure, but you can try to use another kafka_format, eg TSV if there no tabs in data, and extract fields using JSON functions.
```
create table kafka_ingest (
jsonString String,
jsonField1 String DEFAULT JSONExtractString(jsonString, 'field1'),
jsonField2 UInt64 DEFAULT JSONExtractUInt(jsonString, 'field2... | 2019-08-07T12:27:31 | 2019-08-07T12:28:15 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,244,484 | 6,004 | healiseu | Normally in RDBMS optimizers take care of the JOIN order (Left, Right). In the case of ClickHouse, tables with low selectivity at the right side of the JOIN should be preferred and performance will be significantly better. | 2019-08-07T19:51:21 | 2019-08-07T22:15:27 | {} | NONE |
ClickHouse | ClickHouse | 519,028,145 | 6,374 | alexey-milovidov | Need to add a test:
- for the issue;
- to check that the index by integer or decimal continue to work if you convert toFloat64. | 2019-08-07T09:55:07 | 2019-08-07T09:55:07 | {} | MEMBER |
ClickHouse | ClickHouse | 519,119,676 | 6,377 | MeteHanC | > By the way at the moment I am having another issue, when I restart a CH replica under load it does not resume to get queries and act like a cold replica. It resumes to get queries if I restart the CH on the server which my API is connected. Weird.
This is being fixed with #5317 now I think, it is not a bug but ... | 2019-08-07T14:22:34 | 2019-08-07T15:14:35 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,211,523 | 6,386 | alexey-milovidov | Description:
Without this option, Poco library will restart `recv` after `EINTR`, but it doesn't update socket receive timeout that leads to infinite wait when query profiler is activated. The issue persisted in `delayed_replica_failover` integration test after we enabled query profiler by default. | 2019-08-07T18:15:19 | 2019-08-07T18:16:35 | {} | MEMBER |
ClickHouse | ClickHouse | 519,317,770 | 6,355 | heyciao | What I have tested & posted in the previous post was done on one of the PCs.
I have now started a VM on another PC (the same VM on which I could reproduce as well the problem and which I can trash), then I have uninstalled all versions of clang & llvm excluding version 7 (7.1.0) and then I reran the package manager ... | 2019-08-08T00:36:15 | 2019-08-08T00:36:53 | {} | NONE |
ClickHouse | ClickHouse | 519,471,924 | 6,395 | Asura7969 | have already been solved | 2019-08-08T11:00:36 | 2019-08-08T11:00:36 | {} | NONE |
ClickHouse | ClickHouse | 519,580,984 | 6,401 | dgrr | @den-crane when I do `ENGINE = SummingMergeTree(price, spend)` it shows me the following text and a description of how MergeTree works.
```
Code: 42. DB::Exception: Received from localhost:9000. DB::Exception: With extended storage definition syntax storage SummingMergeTree requires 0 to 1 parameters:
list of col... | 2019-08-08T16:00:35 | 2019-08-08T16:02:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,609,508 | 1,424 | zhanglistar | Not yet | 2019-08-06T10:21:33 | 2019-08-06T10:21:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,610,614 | 1,424 | alexey-milovidov | This task is in progress and it is being developed by @vitlibar | 2019-08-06T10:24:58 | 2019-08-06T10:24:58 | {} | MEMBER |
ClickHouse | ClickHouse | 518,699,081 | 6,336 | igor-sh8 | So, in any case 1 topic - it is 1 thread on Clickhouse regardless of the number of partitions in the topic?
Is it possible to configure a few threads on the same topic (maybe one thread per consumer) ? | 2019-08-06T14:40:22 | 2019-08-06T14:40:22 | {} | NONE |
ClickHouse | ClickHouse | 518,909,909 | 6,287 | den-crane | STR:
```
create table radacct (acctstarttime DateTime) Engine=MergeTree order by acctstarttime;
insert into radacct select now() + number%545 from numbers(10000);
insert into radacct select '2019-09-01 00:01:01';
select count() from radacct
where acctstarttime between toDate('2019-09-01') and toDate('2019-09... | 2019-08-07T02:05:52 | 2019-08-07T02:05:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,915,891 | 6,287 | dkhodorchenko | `ENGINE = MergeTree
PARTITION BY toYYYYMM(acctstarttime)
ORDER BY (acctstarttime)
` | 2019-08-07T02:37:50 | 2019-08-07T02:37:50 | {} | NONE |
ClickHouse | ClickHouse | 519,086,672 | 6,334 | astudnev | I think the problem is not related to MergeTree. It is more general. Any queries that use ORDER BY decimal, are executed very very slow | 2019-08-07T13:03:03 | 2019-08-07T13:03:03 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,456,078 | 5,346 | arunk-s | I can confirm that similar behavior is also seen on `ReplicatedCollapsingMergeTree` TableEngine even in CH version `19.4.3` .
| 2019-08-08T10:07:06 | 2019-08-08T10:07:06 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 518,700,054 | 6,338 | alexey-milovidov | @zhang2014 You have pasted articles in a symlink path :rofl: :laughing: that effectively broke my repository (cannot create symlink, file name too long):
```
~/ClickHouse/docs/zh/data_types/domains$ ls -l
total 12
lrwxrwxrwx 1 milovidov dpt_yandex_monetize_metrica_dev_stat 3310 авг 6 17:41 ipv4.md -> '''## IPv4'... | 2019-08-06T14:42:37 | 2019-08-06T14:42:37 | {} | MEMBER |
ClickHouse | ClickHouse | 518,851,659 | 6,341 | den-crane | CH uses 3 tcp ports
8123 -- http
9000 -- tcp/native (clickhouse-client)
9009 -- http replication
make sure that ip:10024 --> 8123 -- http
```
curl 'localhost:8123/?query=select%20event,value%20from%20system.events%20where%20event%3D%27MergedRows%27format%20TSV'
MergedRows 933405
``` | 2019-08-06T21:30:36 | 2019-08-06T21:31:32 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 518,859,020 | 6,355 | heyciao | Thank you for your reply.
The build of "v19.11.5.28-stable" failed.
Here is what I did:
$ git clone --recursive https://github.com/yandex/ClickHouse.git
$ git checkout v19.11.5.28-stable
```
warning: unable to rmdir 'contrib/fastops': Directory not empty
M contrib/hyperscan
M contrib/libunwind
M contrib/po... | 2019-08-06T21:56:27 | 2019-08-06T21:56:27 | {} | NONE |
ClickHouse | ClickHouse | 518,905,823 | 6,357 | yingfeng | Both Doris and Kudu have implemented a state replication machine based infrastructure to resolve distributed issues---through either Paxos or Raft. The solution of Kudu's might be more valuable since it has provided a Raft concensus library in C++ which has also been used by some other distributed OLTP solutions as [Yu... | 2019-08-07T01:44:45 | 2019-08-07T01:44:45 | {} | NONE |
ClickHouse | ClickHouse | 519,069,164 | 6,365 | abyss7 | This scheme looks very promising, especially for Kafka scenarios - but first we must get sure, that it works in our current CI with sane timeouts.
The second thought is that using shell-scripted tests is a step-back after using pytest environment - hopefully `docker-compose` should have some Python bindings. | 2019-08-07T12:11:38 | 2019-08-07T12:11:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 519,189,408 | 6,275 | TH-HA | i'm using the default user in all the case and i've tried to change the settings in multiple way
thru the users.xml / and i can see the changed setting properly set : but same result
thru the clickhouse-client : same result
The log are the following
>
2019.08.07 17:10:05.953491 [ 42 ] {} <Trace> TCPHand... | 2019-08-07T17:15:35 | 2019-08-07T17:16:38 | {} | NONE |
ClickHouse | ClickHouse | 519,258,867 | 6,368 | amchakra789 | Thank you so much, appreciate much your help. I will try and update this thread. | 2019-08-07T20:35:28 | 2019-08-07T20:35:28 | {} | NONE |
ClickHouse | ClickHouse | 519,292,612 | 6,355 | heyciao | Thank you
So, this time the results are mixed:
- on one hand everything seems to work fine (got back in the shell "ok" after executing both "cmake" and "ninja")
- on the other hand I see that at the very end of the compilation, just before that it ends for real, I get the same error message that I mentioned previo... | 2019-08-07T22:32:19 | 2019-08-07T22:32:19 | {} | NONE |
ClickHouse | ClickHouse | 519,381,217 | 6,365 | filimonov | > but first we must get sure, that it works in our current CI with sane timeouts.
In general - there is no tools to monitor Kafka engine behaviour from inside clickhouse. That makes creating good test quite painful.
Some feedback in metrics (like KafakMaximumLag ) / events ( KafkaMessagesConsumed ) can be very... | 2019-08-08T06:14:39 | 2019-08-08T06:31:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 522,575,499 | 6,538 | orivej | odbc-bridge becomes a part of clickhouse-lib here: https://github.com/yandex/ClickHouse/blob/v19.13.2.19-stable/dbms/programs/CMakeLists.txt#L92 (`add_library(clickhouse-lib SHARED … ${CLICKHOUSE_ODBC_BRIDGE_SOURCES})`). | 2019-08-19T13:28:58 | 2019-08-19T13:28:58 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.