Fix bug in cache of the index object client #10587
Merged
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
The bug could manifest in missing results when querying logs older than what is kept on ingesters. This commit changes the behaviour of the caching in the storage client client used by the indexshipper to download indexes from object store. The object store client maintains an internal, in-memory cache of tables (= per-day index directories) available on object storage to reduce the amount of expensive List operations. This cache gets updated in a regular interval (1m) by the table manager that uses the client. However, when the per-tenant setting `query_ready_index_num_days` is set the 0, which is the default (to disable index pre-fetching), the sync is only performed once on startup, but not subsequently. This leads to a stale list of table names in the cache and any new index directory added to the object storage is not added to the client cache. This commit changes the object client to use a read-through caching strategy for listing tables. If a table is not found locally in cache, it performs a lookup against object storage. In this way, the cache is updated even without the regular sync described above. Signed-off-by: Christian Haudum <christian.haudum@gmail.com>
Signed-off-by: Christian Haudum <christian.haudum@gmail.com>
MasslessParticle
approved these changes
Sep 14, 2023
|
Ooof Maybe something went wrong when resolving conflicts after cherry picking the commit. Need to take a look |
2 tasks
rhnasc
pushed a commit
to inloco/loki
that referenced
this pull request
Apr 12, 2024
The bug could manifest in missing results when querying logs older than what is kept on ingesters. This commit changes the behaviour of the caching in the storage client client used by the indexshipper to download indexes from object store. The object store client maintains an internal, in-memory cache of tables (= per-day index directories) available on object storage to reduce the amount of expensive List operations. This cache gets updated in a regular interval (1m) by the table manager that uses the client. However, when the per-tenant setting `query_ready_index_num_days` is set the 0, which is the default (to disable index pre-fetching), the sync is only performed once on startup, but not subsequently. This leads to a stale list of table names in the cache and any new index directory added to the object storage is not added to the client cache. This commit changes the object client to use a read-through caching strategy for listing tables. If a table is not found locally in cache, it performs a lookup against object storage. In this way, the cache is updated even without the regular sync described above. **Special notes for your reviewer**: This is a port of grafana#10585 which was first applied on the 2.9 release branch. --------- Signed-off-by: Christian Haudum <christian.haudum@gmail.com>
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.
The bug could manifest in missing results when querying logs older than what is kept on ingesters.
This commit changes the behaviour of the caching in the storage client client used by the indexshipper to download indexes from object store.
The object store client maintains an internal, in-memory cache of tables (= per-day index directories) available on object storage to reduce the amount of expensive List operations. This cache gets updated in a regular interval (1m) by the table manager that uses the client. However, when the per-tenant setting
query_ready_index_num_daysis set the 0, which is the default (to disable index pre-fetching), the sync is only performed once on startup, but not subsequently. This leads to a stale list of table names in the cache and any new index directory added to the object storage is not added to the client cache.This commit changes the object client to use a read-through caching strategy for listing tables. If a table is not found locally in cache, it performs a lookup against object storage. In this way, the cache is updated even without the regular sync described above.
Special notes for your reviewer:
This is a port of #10585 which was first applied on the 2.9 release branch.