In #33291 final part commit had been defered, and now it can take
significantly more time, that may lead to "Part directory doesn't exist"
error during INSERT:
2022.02.21 18:18:06.979881 [ 11329 ] {insert} <Debug> executeQuery: (from 127.1:24572, user: default) INSERT INTO db.table (...) VALUES
2022.02.21 20:58:03.933593 [ 11329 ] {insert} <Trace> db.table: Renaming temporary part tmp_insert_20220214_18044_18044_0 to 20220214_270654_270654_0.
2022.02.21 21:16:50.961917 [ 11329 ] {insert} <Trace> db.table: Renaming temporary part tmp_insert_20220214_18197_18197_0 to 20220214_270689_270689_0.
...
2022.02.22 21:16:57.632221 [ 64878 ] {} <Warning> db.table: Removing temporary directory /clickhouse/data/db/table/tmp_insert_20220214_18232_18232_0/
...
2022.02.23 12:23:56.277480 [ 11329 ] {insert} <Trace> db.table: Renaming temporary part tmp_insert_20220214_18232_18232_0 to 20220214_273459_273459_0.
2022.02.23 12:23:56.299218 [ 11329 ] {insert} <Error> executeQuery: Code: 107. DB::Exception: Part directory /clickhouse/data/db/table/tmp_insert_20220214_18232_18232_0/ doesn't exist. Most likely it is a logical error. (FILE_DOESNT_EXIST) (version 22.2.1.1) (from 127.1:24572) (in query: INSERT INTO db.table (...) VALUES), Stack trace (when copying this message, always include the lines below):
Follow-up for: #28760
Refs: #33291
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
fsync of the temporary part directory is superfluous anyway, and besides
that directory is not exists at that time, that will lead to ENOENT
error:
2022.02.18 17:02:51.634565 [ 35639 ] {} <Error> void DB::MergeTreeBackgroundExecutor<DB::MergeMutateRuntimeQueue>::routine(DB::TaskRuntimeDataPtr) [Queue = DB::MergeMutateRuntimeQueue]: Code: 107. DB::ErrnoException: Cannot open file /var/lib/clickhouse/data/system/text_log/tmp_merge_202202_1864_3192_14/, errno: 2, strerror: No such file or directory. (FILE_DOESNT_EXIST), Stack trace (when copying this message, always include the lines below):
0. DB::Exception::Exception() @ 0xb26ecfa in /usr/lib/debug/.build-id/01/8c328bd4858d67.debug
1. DB::throwFromErrnoWithPath() @ 0xb2700ea in /usr/lib/debug/.build-id/01/8c328bd4858d67.debug
2. DB::LocalDirectorySyncGuard::LocalDirectorySyncGuard() @ 0x14905531 in /usr/lib/debug/.build-id/01/8c328bd4858d67.debug
3. DB::DiskLocal::getDirectorySyncGuard() const @ 0x148af3e3 in /usr/lib/debug/.build-id/01/8c328bd4858d67.debug
4. DB::MergeTask::ExecuteAndFinalizeHorizontalPart::prepare() @ 0x157bef13 in /usr/lib/debug/.build-id/01/8c328bd4858d67.debug
Note, that IMergeTreeDataPart::renameTo() anyway will have fsync for the
directory.
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
There are two possible cases for execution merges/mutations:
1) from background thread
2) from OPTIMIZE TABLE query
1) is pretty simple, it's memory tracking structure is as follow:
current_thread::memory_tracker = level=Thread / description="(for thread)" ==
background_thread_memory_tracker = level=Thread / description="(for thread)"
current_thread::memory_tracker.parent = level=Global / description="(total)"
So as you can see it is pretty simple and MemoryTrackerThreadSwitcher
does not do anything icky for this case.
2) is complex, it's memory tracking structure is as follow:
current_thread::memory_tracker = level=Thread / description="(for thread)"
current_thread::memory_tracker.parent = level=Process / description="(for query)" ==
background_thread_memory_tracker = level=Process / description="(for query)"
Before this patch to track memory (and related things, like sampling,
profiling and so on) for OPTIMIZE TABLE query dirty hacks was done to
do this, since current_thread memory_tracker was of Thread scope, that
does not have any limits.
And so if will change parent for it to Merge/Mutate memory tracker
(which also does not have some of settings) it will not be correctly
tracked.
To address this Merge/Mutate was set as parent not to the
current_thread memory_tracker but to it's parent, since it's scope is
Process with all settings.
But that parent's memory_tracker is the memory_tracker of the
thread_group, and so if you will have nested ThreadPool inside
merge/mutate (this is the case for s3 async writes, which has been
added in #33291) you may get use-after-free of memory_tracker.
Consider the following example:
MemoryTrackerThreadSwitcher()
thread_group.memory_tracker.parent = merge_list_entry->memory_tracker
(see also background_thread_memory_tracker above)
CurrentThread::attachTo()
current_thread.memory_tracker.parent = thread_group.memory_tracker
CurrentThread::detachQuery()
current_thread.memory_tracker.parent = thread_group.memory_tracker.parent
# and this is equal to merge_list_entry->memory_tracker
~MemoryTrackerThreadSwitcher()
thread_group.memory_tracker = thread_group.memory_tracker.parent
So after the following we will get incorrect memory_tracker (from the
mege_list_entry) when the next job in that ThreadPool will not have
thread_group, since in this case it will not try to update the
current_thread.memory_tracker.parent and use-after-free will happens.
So to address the (2) issue, settings from the parent memory_tracker
should be copied to the merge_list_entry->memory_tracker, to avoid
playing with parent memory tracker.
Note, that settings from the query (OPTIMIZE TABLE) is not available at
that time, so it cannot be used (instead of parent's memory tracker
settings).
v2: remove memory_tracker.setOrRaiseHardLimit() from settings
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
It is possible to get empty part in future parts, during merging, that
can be created in case of there is information about this part in
zookeeper, but no replicas has it
(StorageReplicatedMergeTree::createEmptyPartInsteadOfLost())
And this may incorrectly initialize min value, which will be updated
after one more merge (since during that merge there will be no more
information about empty part) or OPTIMIZE.
