batcheval: keep correct stats in AddSSTable without overlap #36907
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
Clarify why it's OK to ingest non-atomically with applying the Raft command. Release note: None
Release note: None
|
This change is |
dt
approved these changes
Apr 17, 2019
dt
reviewed
Apr 17, 2019
pkg/storage/replica_raft.go
Outdated
| // | ||
| // Note that in the event of an ill-timed crash, the data may have been | ||
| // ingested (i.e. become "visible") without the Raft command having | ||
| // marked as applied. However, the node will only serve reads once a new |
There was a problem hiding this comment.
Follower reads will see extra data, which is weird (and the data could be out-of-order with the raft log...) anyway, hard to figure what exactly a problem here would be, but closed timestamps would definitely go down the drain and everything is pretty undefined with sst ingestion anyway. I rearranged the comment and clarified.
For distributed index backfill, we recently switched to allowing the stats of the data in the SST to be blindly added to that of the range (thus introducing ContainsEstimates=true). However, if the keyspace being ingested into is empty (such as usually the case during IMPORT/RESTORE), it seems a little silly to make all the consistency checks for the next 24h twice as expensive. Probe whether the keyspace is empty and if so, don't pretend to introduce an approximation - the stats in this case are exact. Release note: None
Release note: None
|
Superseded by many changes in cmd_add_sstable.go that have landed since. |
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Outdated suggestions cannot be applied.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.
For distributed index backfill, we recently switched to allowing the stats
of the data in the SST to be blindly added to that of the range
(thus introducing ContainsEstimates=true). However, if the keyspace being
ingested into is empty (such as usually the case during IMPORT/RESTORE), it
seems a little silly to make all the consistency checks for the next 24h
twice as expensive. Probe whether the keyspace is empty and if so, don't
pretend to introduce an approximation - the stats in this case are exact.
Release note: None