Make LRUQueryCache respect Accountable queries on eviction and consisten… #12614
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
romseygeek
requested changes
Oct 3, 2023
| @@ -385,7 +385,9 @@ public void clearQuery(Query query) { | |||
|
|
|||
| private void onEviction(Query singleton) { | |||
| assert lock.isHeldByCurrentThread(); | |||
| onQueryEviction(singleton, LINKED_HASHTABLE_RAM_BYTES_PER_ENTRY + QUERY_DEFAULT_RAM_BYTES_USED); | |||
| var ramBytesUsedByQuery = LINKED_HASHTABLE_RAM_BYTES_PER_ENTRY; | |||
| ramBytesUsedByQuery += singleton instanceof Accountable accountableQuery ? accountableQuery.ramBytesUsed() : QUERY_DEFAULT_RAM_BYTES_USED; | |||
There was a problem hiding this comment.
Can we pull this out into a separate queryRamBytesUsed(Query q) method, given that it's used twice?
|
Can you run |
gtroitskiy
force-pushed
the
fix_query_cache
branch
from
987d564
to
9b96da8
Compare
|
Thanks for reviewing! I ran tidy and made some refactoring |
romseygeek
reviewed
Oct 3, 2023
gtroitskiy
force-pushed
the
fix_query_cache
branch
from
9b96da8
to
ec6c3fd
Compare
gtroitskiy
changed the title
romseygeek
approved these changes
Oct 3, 2023
romseygeek
pushed a commit
that referenced
this pull request
Oct 3, 2023
…tency check (#12614) Given a query that implements Accountable, the LRUQueryCache would increment its internal accounting by the amount reported by Accountable.ramBytesUsed(), but only decrement on eviction by the default used for all other queries. This meant that the cache could eventually think it had run out of space, even if there were no queries in it at all. This commit ensures that queries that implement Accountable are always accounted for correctly.
s1monw
pushed a commit
to s1monw/lucene
that referenced
this pull request
Oct 10, 2023
…tency check (apache#12614) Given a query that implements Accountable, the LRUQueryCache would increment its internal accounting by the amount reported by Accountable.ramBytesUsed(), but only decrement on eviction by the default used for all other queries. This meant that the cache could eventually think it had run out of space, even if there were no queries in it at all. This commit ensures that queries that implement Accountable are always accounted for correctly.
|
I am also facing the same issue after the upgrade of elastic to 8.10 from 8.7. |
|
@gtroitskiy @romseygeek |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
…cy check
Root cause
onQueryCacheincreasesramBytesUsedfor specified amount, that is being calculated with respect to query beingAccountableor not.Unfortunately,
onQueryEvictiondoes not the same. If some heavy accountable query hasramBytesUsed()greater thanQUERY_DEFAULT_RAM_BYTES_USED, the deltaghostBytes = ramBytesUsed() - QUERY_DEFAULT_RAM_BYTES_USEDremains in totalLRUQueryCache#ramBytesUsedforever.Hit rate drops to 0
since total sum of ghost bytes monotonously increases with each cached accountable query (>QUERY_DEFAULT_RAM_BYTES_USED), eventually it becomes greater than
maxRamBytesUsed, and each newly cached query immediately evicts from the cache.Current behavior of LRUQueryCache for some real service in production [though not fully optimized]
maxRamBytesUsed. Started eviction. From now on size of cached queries decreases to compensate increasing ghost bytes.maxRamBytesUsed. If any new query was cached, it evicted at the same time. Cache size is 0. Hit rate is 0.After fix
