Round robin fixes #1336
Closed
Round robin fixes #1336
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
Added a test which fails: creating two iterators and iterating in order ABAB would return two copies of node-1 for A, two copies of node-2 for B. This was pretty useless, since each iterator shouldn't return the same node twice, and should return all nodes eventually. The bug was that the index used to return the next node was global and shared between all iterators! The code change causes the new test to pass.
… any remote hosts at all! Added tests to cover missing cases and fix code. The old code simply never fell back to remote entries after exhausting local nodes. The new code does round-robin over local, then when it runs out of local, does round-robin over remote nodes. We also don't return remote nodes for LOCAL_ONE and LOCAL_QUORUM queries, to mimic the datastax Java driver's behaviour.
NWilson
force-pushed
the
ncw/round-robin-fixes
branch
from
4a31657
to
f964952
Compare
alourie
requested changes
Aug 5, 2019
| } | ||
|
|
||
| hosts := r.hosts.get() | ||
| var startIdx int |
There was a problem hiding this comment.
I'd think it would be cleaner here to return nil immediately if len(hosts) == 0
There was a problem hiding this comment.
That's a stylistic change I'd be happy to do.
But are you sure? None of the existing Pick() implementations returns nil - the function they return can return nil, but the NextHost function itself is always non-nil.
I tried making the change, but one of the tests currently asserts that Pick() returns a non-nil function even when there are no hosts.
There was a problem hiding this comment.
Uhm, ok, that's a valid point. Ignore then ;-)
alourie
approved these changes
Aug 18, 2019
| } | ||
|
|
||
| hosts := r.hosts.get() | ||
| var startIdx int |
There was a problem hiding this comment.
Uhm, ok, that's a valid point. Ignore then ;-)
|
fixed via #1351 |
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.
There were some nasty issues with the round-robin host policies!
The round-robin host policy wasn't actually returning all the nodes to each iterator, you'd get queries where it would try some nodes twice, some not at all, because the iterators were using a shared index.
The DC-aware round-robin policy was similarly broken, and wouldn't even return any remote nodes: as long as there were any local nodes, it would never query a remote node, so if you were in a cluster where the local nodes went down the driver wouldn't fall back to anything else!
I've closely copied the strategy used by the Datastax Java driver, so I think these fixes are right.
Added tests (which fail without the code changes). See individual commit messages for further details.