This repository has been archived by the owner on Nov 7, 2023. It is now read-only.
Fix prepare/request timers #142
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
nhoriguchi
reviewed
Dec 15, 2019
core/commit.go
Outdated
|
|
||
| // FIXME: Only stop the request timer if the request | ||
| // identifier it was started for matches the one of | ||
| // the request accepted for execution. |
There was a problem hiding this comment.
I agree with this, thank you for spotting.
There was a problem hiding this comment.
Is there a possibility that a timer is left without stopping forever? I am not sure how a timer is stopped certainly at some point if stopReqTimer() doesn't stop the timer at this point.
There was a problem hiding this comment.
Since correct clients only send their requests sequentially, a replica need to maintain one prepare and one request timer per client. If a newer request is received, the replica can assume the former request was completed and the client timers should be reused. In other words, replicas only maintain timers for the most recent request of each client.
Merged
Similar to prepareTimerStarter, change definition of prepareTimerStopper to receive a Request message as parameter. Signed-off-by: Sergey Fedorov <sergey.fedorov@neclab.eu>
added 4 commits
December 17, 2019 22:23
The purpose of the prepare timer is to trigger relaying of the Request message to the primary. So there is no point in starting prepare timer when the Request is handled by the primary replica itself. Signed-off-by: Sergey Fedorov <sergey.fedorov@neclab.eu>
Signed-off-by: Sergey Fedorov <sergey.fedorov@neclab.eu>
Similar to prepareTimerStarter/prepareTimerStopper, pass Request message to requestTimerStarter/requestTimerStopper. This makes the both timers' abstractions similar to each other. As opposed to passing just client identifier, this will allow to check request identifier without further changing those definitions. Signed-off-by: Sergey Fedorov <sergey.fedorov@neclab.eu>
It may happen that a replica receives a newer client request before it has received Prepare for a former request from that client or accepts such request for execution. In that case, request and prepare timers started for the newer request should not be stopped. Tag request/prepare timers with a request identifier when started and stop the timer only when the same request identifier supplied. Signed-off-by: Sergey Fedorov <sergey.fedorov@neclab.eu>
sergefdrv
force-pushed
the
fix-prepare-timer
branch
from
5066756 to
eeeb504
Compare
|
Summary of changes:
|
nhoriguchi
approved these changes
Dec 17, 2019
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
This pull request fixes some issues with prepare timer that were missed when reviewing #127.
Depending on how we decide to solve the issue noted in "WIP: core: Check request ID before stopping request/prepare timer", we might decide to do the opposite of what "RFC: core: Pass client ID to prepareTimerStarter" suggests.