Execute DeleteRetentionPolicy and DeleteMeasurement in parallel
#6847
Conversation
|
By analyzing the blame information on this pull request, we identified @e-dard and @joelegasse to be potential reviewers |
|
/cc @jwilder There's one caveat for something I saw in |
jsternberg
force-pushed
the
js-parallel-database-deletes
branch
3 times, most recently
from
84caca7 to
8939d18
Compare
| return err | ||
| return func() error { | ||
| // Delete the shard from disk. | ||
| return s.DeleteShard(shardID) |
There was a problem hiding this comment.
This is needed to avoid a data race, but I'm not sure how useful this is other than just ensuring the database doesn't get locked up while a retention policy is being removed from shards. This takes a lock on the tsdb.Store instance so there's no way that this can be run in parallel.
jsternberg
force-pushed
the
js-parallel-database-deletes
branch
from
8939d18 to
4422102
Compare
| s.mu.Lock() | ||
| defer s.mu.Unlock() | ||
| m, err := func() (*Measurement, error) { | ||
| s.mu.Lock() |
There was a problem hiding this comment.
This can be a read-lock because the indexes have their own mutexs. It should be released after 492 too. Taking a write lock here will lock writes and queries for all databases until this completes which is not necessary.
Makes the code from `DeleteDatabase` that performed the parallel delete more general and reuses it for deleting a retention policy and a measurement.
jsternberg
force-pushed
the
js-parallel-database-deletes
branch
from
4422102 to
6da3162
Compare
|
Fixed via #7015 |
Makes the code from
DeleteDatabasethat performed the parallel deletemore general and reuses it for deleting a retention policy and a
measurement.