After detachQueryIfNotDetached() had been removed it is not enough to
use attachTo() for ThreadPool (scheduleOrThrowOnError()) since the query
may be already attached, if the thread doing multiple jobs, so
CurrentThread::attachToIfDetached() should be used instead.
This should fix all the places from the failures on CI [1]:
$ fgrep DB::CurrentThread::attachTo -A1 ~/Downloads/47.txt | fgrep -v attachTo | cut -d' ' -f5,6 | sort | uniq -c
92 --
2 /fasttest-workspace/build/../../ClickHouse/contrib/libcxx/include/deque:1393: DB::ParallelParsingInputFormat::parserThreadFunction(std::__1::shared_ptr<DB::ThreadGroupStatus>,
4 /fasttest-workspace/build/../../ClickHouse/src/Storages/MergeTree/MergeTreeData.cpp:1595: void
87 /fasttest-workspace/build/../../ClickHouse/src/Storages/MergeTree/MergeTreeDataSelectExecutor.cpp:993: void
[1]: https://github.com/ClickHouse/ClickHouse/runs/4954466034?check_suite_focus=true
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
MemoryTracker starts accounting memory directly only after per-thread
allocation exceeded max_untracker_memory (or memory_profiler_step).
But even memory under this limit should be accounted too, and there is
code to do this in ThreadStatus dtor, however due to
PullingAsyncPipelineExecutor detached the query from thread group that
memory was not accounted.
So remove CurrentThread::detachQueryIfNotDetached() from threads that
uses ThreadFromGlobalPool since it has ThreadStatus, and the query will
be detached using CurrentThread::defaultThreadDeleter.
Note, that before this patch memory accounting works for HTTP queries
due to it had been accounted from ParallelFormattingOutputFormat, but
not for TCP.
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
Before it was possible to return stale counter from
StorageMergeTree::totalRows() (that is used for
optimize_trivial_count_query) since the accounting is done in two steps:
- subtract old number of rows
<-- here the number can be zero, even though there are rows -->
- add new number of rows
This was found by CI [1] in 01615_random_one_shard_insertion test:
Here you can see that INSERT went to both tables:
<details>
2022.01.16 09:07:34.288252 [ 154369 ] {a1905be0-93da-460c-8c6f-9b5adace72a0} <Debug> DistributedBlockOutputStream: It took 0.035197041 sec. to insert 100
blocks, 2841.1479249065287 rows per second. Insertion status:
Wrote 54 blocks and 54 rows on shard 0 replica 0, localhost:9000 (average 0 ms per block, the slowest block 1 ms)
Wrote 46 blocks and 46 rows on shard 1 replica 0, localhost:9000 (average 0 ms per block, the slowest block 1 ms)
</details>
But the test fails, since select from shard1.tbl returns 0, and the
problem was concurrent merge:
<details>
2022.01.16 09:07:34.289470 [ 146495 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Debug> executeQuery: (from [::1]:42082) (comment: 01615_random_one_shard_insertion.sql) select count() != 0 from shard_0.tbl;
2022.01.16 09:07:34.289564 [ 375 ] {c7a885fa-4ef4-4dcf-a4de-1650d44fa0ab::all_1_54_9} <Debug> MergeTask::MergeProjectionsStage: Merge sorted 54 rows, containing 1 columns (1 merged, 0 gathered) in 0.00171193 sec., 31543.345814373253 rows/sec., 246.43 KiB>
2022.01.16 09:07:34.289810 [ 375 ] {c7a885fa-4ef4-4dcf-a4de-1650d44fa0ab::all_1_54_9} <Trace> shard_0.tbl (c7a885fa-4ef4-4dcf-a4de-1650d44fa0ab): Renaming temporary part tmp_merge_all_1_54_9 to all_1_54_9.
2022.01.16 09:07:34.289858 [ 146495 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Trace> ContextAccess (default): Access granted: SELECT(number) ON shard_0.tbl
2022.01.16 09:07:34.289897 [ 375 ] {c7a885fa-4ef4-4dcf-a4de-1650d44fa0ab::all_1_54_9} <Trace> shard_0.tbl (c7a885fa-4ef4-4dcf-a4de-1650d44fa0ab) (MergerMutator): Merged 6 parts: from all_1_49_8 to all_54_54_0
2022.01.16 09:07:34.289920 [ 146495 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Trace> InterpreterSelectQuery: WithMergeableState -> Complete
2022.01.16 09:07:34.289987 [ 375 ] {} <Debug> MemoryTracker: Peak memory usage Mutate/Merge: 3.12 MiB.
2022.01.16 09:07:34.290305 [ 154344 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Trace> MergingAggregatedTransform: Reading blocks of partially aggregated data.
2022.01.16 09:07:34.290332 [ 154344 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Debug> MergingAggregatedTransform: Read 1 blocks of partially aggregated data, total 1 rows.
2022.01.16 09:07:34.290343 [ 154344 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Trace> Aggregator: Merging partially aggregated single-level data.
2022.01.16 09:07:34.290358 [ 154344 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Trace> Aggregator: Merged partially aggregated single-level data.
2022.01.16 09:07:34.290366 [ 154344 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Trace> Aggregator: Converting aggregated data to blocks
2022.01.16 09:07:34.290391 [ 154344 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Debug> Aggregator: Converted aggregated data to blocks. 1 rows, 8.00 B in 1.0939e-05 sec. (91416.034 rows/sec., 714.19 KiB/sec.)
2022.01.16 09:07:34.290709 [ 146495 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Information> executeQuery: Read 1 rows, 4.01 KiB in 0.001187722 sec., 841 rows/sec., 3.30 MiB/sec.
2022.01.16 09:07:34.290774 [ 146495 ] {cd9d4cf2-7131-4179-b0b2-3aeec4045755} <Debug> MemoryTracker: Peak memory usage (for query): 0.00 B.
</details>
[1]: https://s3.amazonaws.com/clickhouse-test-reports/33675/7848ea7d609e4c720e8e4494eb6207c0751f5aea/stateless_tests__ubsan__actions_.html
This also fixes a race between DROP TABLE check and INSERT/merges.
v0: use Active parts instead.
v2: fix total counters accounting instead.
