-
Notifications
You must be signed in to change notification settings - Fork 3.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
kvserver: use leader leases in various flow control tests #136182
kvserver: use leader leases in various flow control tests #136182
Conversation
This test sets a very large number for RaftElectionTimeoutTicks to keep leadership sticky. Naively, this prevents us from using leader leases in this test, as replicas won't campaign if they don't have StoreLiveness support, which isn't kicked off until the first tick. Conveniently, raft fortification is yet another way to keep leadership sticky -- so explicitly using leader leases should achieve the same goal. References cockroachdb#133763 Release note: None
This test sets a very large number for RaftElectionTimeoutTicks to keep leadership sticky. Naively, this prevents us from using leader leases in this test, as replicas won't campaign if they don't have StoreLiveness support, which isn't kicked off until the first tick. Conveniently, raft fortification is yet another way to keep leadership sticky -- so explicitly using leader leases should achieve the same goal. References cockroachdb#133763 Release note: None
This test sets a very large number for RaftElectionTimeoutTicks to keep leadership sticky. Naively, this prevents us from using leader leases in this test, as replicas won't campaign if they don't have StoreLiveness support, which isn't kicked off until the first tick. Conveniently, raft fortification is yet another way to keep leadership sticky -- so explicitly using leader leases should achieve the same goal. References cockroachdb#133763 Release note: None
This test sets a very large number for RaftElectionTimeoutTicks to keep leadership sticky. Naively, this prevents us from using leader leases in this test, as replicas won't campaign if they don't have StoreLiveness support, which isn't kicked off until the first tick. Conveniently, raft fortification is yet another way to keep leadership sticky -- so explicitly using leader leases should achieve the same goal. References cockroachdb#133763 Release note: None
We'll soon enable leader leases metamorphically. There's a few tests that want to always use a lease type that allows for quiescence -- so instead of disabling expiration based leases, they should instead explicitly set its lease type to be epoch based. The tests are: - TestFlowControlUnquiescedRangeV2 - TestFlowControlUnquiescedRange - TestFlowControlQuiescedRange References cockroachdb#133763 Release note: None
Much like the previous commits, a bunch of other tests were setting a very large number for RaftElectionTimeoutTicks to keep leadership sticky. Naively, this prevents us from using leader leases in these tests, as replicas won't campaign unless they have StoreLiveness support, which they won't until the first tick kicks off a request. Conveniently, raft fortification is yet another way to keep leadership sticky -- so explicitly using leader leases should achieve the same goal. The tests modified here are: - TestFlowControlSendQueue - TestFlowControlRepeatedlySwitchMode - TestFlowControlRaftSnapshot - TestFlowControlRaftTransportBreak - TestFlowControlAdmissionPostSplitMergeV2 - TestFlowControlV1ToV2Transition References cockroachdb#133763 Release note: None
At one point, I got tired of pulling things out into their own commits and decided to batch. I was too lazy to go back and batch the first few commits, so unless you (@kvoli, @iskettaneh) care about the lack of uniformity, here we are. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No testing diffs in .../kvserver/testdata/flow_control_integration_v2/*
+ passing tests is good indication that these changes had little discernible impact on rac2.
Reviewed 1 of 1 files at r6, all commit messages.
Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @iskettaneh)
TFTR! bors r=kvoli |
See individual commits for details.