Stream Raft Messages and Fix Check Quorum #3138
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
martinmr
suggested changes
Mar 15, 2019
There was a problem hiding this comment.
Reviewed 3 of 4 files at r1, 1 of 1 files at r2.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @manishrjain)
conn/node.go, line 379 at r2 (raw file):
/ Exit after a thousand tries
"Exit after at least a thousand tries or ten seconds" to match what the code is actually doing.
conn/node.go, line 386 at r2 (raw file):
So that we print error only a few times
nit: "Update lastLog so that we print the error ... "
manishrjain
commented
Mar 20, 2019
There was a problem hiding this comment.
Reviewable status: 1 of 5 files reviewed, 2 unresolved discussions (waiting on @martinmr)
conn/node.go, line 379 at r2 (raw file):
Previously, martinmr (Martin Martinez Rivera) wrote…
/ Exit after a thousand tries"Exit after at least a thousand tries or ten seconds" to match what the code is actually doing.
Done.
conn/node.go, line 386 at r2 (raw file):
Previously, martinmr (Martin Martinez Rivera) wrote…
So that we print error only a few timesnit: "Update lastLog so that we print the error ... "
Done.
manishrjain
changed the title
manishrjain
added a commit
that referenced
this pull request
Mar 20, 2019
Instead of sending the Raft messages, one message per gRPC call, this PR creates a one-way stream between the sender and the receiver. Each messages gets pushed to a channel. We use smart batching to pick up as many messages as we can and send them over the stream in order. If we see connection issues etc., there are mechanisms in place to recreate the stream. Another issue I saw was related to Zero being unable to maintain quorum. It was because of an unbuffered channel in checkQuorum asking for read index, which didn't allow multiple requests to be pushed into one batch causing check quorum to fail even with one second timeout. After allocating a buffered channel, all the check quorum requests finish within a millisecond, rarely going above 7ms in my tests. Changes: * Stream raft messages instead of sending them one by one. * Set duration to 10s * Zero checkQuorum works well now * Martin's comments * Adjust timeouts in contexts, so the deeper one has shorter timeout and the outer one has longer. * Batch up multiple Raft messages from channel and send them in one request.
dna2github
pushed a commit
to dna2fork/dgraph
that referenced
this pull request
Jul 19, 2019
Instead of sending the Raft messages, one message per gRPC call, this PR creates a one-way stream between the sender and the receiver. Each messages gets pushed to a channel. We use smart batching to pick up as many messages as we can and send them over the stream in order. If we see connection issues etc., there are mechanisms in place to recreate the stream. Another issue I saw was related to Zero being unable to maintain quorum. It was because of an unbuffered channel in checkQuorum asking for read index, which didn't allow multiple requests to be pushed into one batch causing check quorum to fail even with one second timeout. After allocating a buffered channel, all the check quorum requests finish within a millisecond, rarely going above 7ms in my tests. Changes: * Stream raft messages instead of sending them one by one. * Set duration to 10s * Zero checkQuorum works well now * Martin's comments * Adjust timeouts in contexts, so the deeper one has shorter timeout and the outer one has longer. * Batch up multiple Raft messages from channel and send them in one request.
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.
Instead of sending the Raft messages, one message per gRPC call, this PR creates a one-way stream between the sender and the receiver. Each messages gets pushed to a channel. We use smart batching to pick up as many messages as we can and send them over the stream in order. If we see connection issues etc., there are mechanisms in place to recreate the stream.
Another issue I saw was related to Zero being unable to maintain quorum. It was because of an unbuffered channel in checkQuorum asking for read index, which didn't allow multiple requests to be pushed into one batch causing check quorum to fail even with one second timeout. After allocating a buffered channel, all the check quorum requests finish within a millisecond, rarely going above 7ms in my tests.
This change is