RFC:follower read cache #116
Conversation
|
Welcome @mittalrishabh! It looks like this is your first PR to tikv/rfcs 🎉 |
text/0113-follower-read-cache.md
Outdated
| Changes to advance safe ts from raft read index response is pretty straightforward and most of the existing logic in the stale read path will be reused. | ||
| Ref counters per region need to be added in the leader to keep track of total number of pending locks and lock status need to be included in a raft read index response. These are the detailed steps | ||
| - Read index request check the lock reference counter and send its status through raft message in read index response | ||
| - Update safe ts if there are no locks at this timestamp in replica. |
There was a problem hiding this comment.
Why should locks on replicas be considered? For advancing resolved-ts or safe-ts, the current TiKV logic determines whether to update safe-ts based on the timestamp and applied index sent by the leader.
Perhaps the safe-ts update should reuse existing processes to ensure correctness.
There was a problem hiding this comment.
We are going to use the existing logic to update safe-ts. It is raft who is going to update safe-ts only when there are no locks on leader.
|
Thanks for your pull request. Before we can look at it, you'll need to add a 'DCO signoff' to your commits. 📝 Please follow instructions in the contributing guide to update your commits with the DCO Full details of the Developer Certificate of Origin can be found at developercertificate.org. The list of commits missing DCO signoff:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
RFC to improve the performance and reliability of follower reads.