logstore: load last entry term #142016
Draft
logstore: load last entry term #142016
+79
−79
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
See the function comment. RawNode never accesses log indices above the storage.LastIndex() because it has them covered by the unstable log. Epic: none Release note: none
Epic: none Release note: none
|
This change is |
|
Stacked on top of #142014. |
We already read the last entry term when initializing RawNode, so there is no additional overhead, except the fact that this entry is now not stashed into the raft entry cache. There appears to be low benefit in putting it into the cache though. Typically, we wouldn't need exactly the LastIndex entry post startup. Eventually, on startup we will load the term cache instead of the last entry. Epic: none Release note: none
We always know the term of the last entry. It is now loaded on replica startup. The applySnapshot code resets the log to empty, and knows the term of the last entry. Epic: none Release note: none
pav-kv
force-pushed
the
load-last-term
branch
from
47128f4 to
0d6c315
Compare
pav-kv
commented
Feb 26, 2025
Comment on lines
738
to
+739
| LastIndex: state.RaftAppliedIndex, | ||
| LastTerm: invalidLastTerm, | ||
| LastTerm: state.RaftAppliedIndexTerm, |
There was a problem hiding this comment.
Logically, these should be set based on truncState (scoped to log storage) rather than ReplicaState (scoped to state machine), even though these index/term are the same in applySnapshot (need to double-check).
There was a problem hiding this comment.
Seems to involve a lot of refactor, lets save this task for the future.
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Outdated suggestions cannot be applied.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.
We already read the last entry term when initializing
RawNode, so there is no additional overhead, except the fact that this entry is now not stashed into the raft entry cache. There appears to be low benefit in putting it into the cache though. Typically, we wouldn't need exactly theLastIndexentry post startup, it would more likely be an earlier entry. So puttingLastIndexinto the cache actually makes it ineffective because the cache drops entries below the cached slice.Eventually, on startup we will load the term cache instead of the last entry.
As a nice side-effect, this PR removes the
invalidLastTermoddity. It also introduces an invariant thatReplicaalways knows the last raft log entry index/term. We can use this to improve theraft.StorageAPI (e.g. addLastEntryID()method and remove this panic).Resolves #142013