Persist txn data listen next simplify #27134
Closed
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 `TxnsCache` and `TxnsCacheState` structs are in-memory caches of
the persist txn shard. It allows callers to ask questions about
specific shards at a specific timestamp. It also allows callers to
compact the cache up to some timestamp.
Previously, the compaction could compact the cache past unapplied
writes. Then, when a caller would ask certain questions, such as "what
was the most recent previous unapplied write at ts?", the cache might
incorrectly return that there were no unapplied writes since any
information of them was compacted away. Specifically, this was
happening in the `latest_write` and `empty_to` return values of the
`data_snapshot` method.
This commit fixes the issue by removing the user facing API for
compacting the cache. Instead, the cache is self compacting such that
it maintains the following invariants:
- All unapplied writes, registers, and forgets are included in the
cache.
- The latest write and registration are always kept in the cache.
While we're changing `data_snapshot`, we remove the assumption that
forgets apply all previous writes. `TxnsCacheState` has no way of
enforcing this invariant and relies on callers to enforce it. It also
has no way of validating the invariant. This makes it potentially scary
for `TxnsCacheState` to rely on this assumption and easy to
accidentally break code that relies on the assumption. If callers want
the cache to take advantage of forgets applying previous writes, then
they should (and do) update the cache state to reflect this. Then the
`TxnsCacheState` can look only at its current state to learn about
applied vs unapplied writes without making any assumptions. A
consequence of this is that the cache must wait for a write to be
applied **and tidied** before learning that it's applied. This is maybe
a slight performance regression in some specific cases, but should
hopefully save some head scratching. If we need to in the future, we
can have callers remove writes from the cache after applying them but
before tidying them.
Fixes MaterializeInc#26893
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.
Motivation
Tips for reviewer
Checklist
$T ⇔ Proto$Tmapping (possibly in a backwards-incompatible way), then it is tagged with aT-protolabel.