-
Notifications
You must be signed in to change notification settings - Fork 2.1k
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
WIP: vitess Online DDL atomic cut-over #11418
WIP: vitess Online DDL atomic cut-over #11418
Conversation
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Review ChecklistHello reviewers! 👋 Please follow this checklist when reviewing this Pull Request. General
If a new flag is being introduced:
If a workflow is added or modified:
Bug fixes
Non-trivial changes
New/Existing features
Backward compatibility
|
…ES statement Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
Signed-off-by: Shlomi Noach <[email protected]>
So far no good results. Hammering the branch against |
Superseded by #11460, which takes a stricter and safer approach. |
Description
This is work in progress for an atomic
RENAME
(not two-step gaping hole) in the cut-over process of a vitess/vreplication Online DDL migration.Curently, we run a two-step rename, when the original table is renamed away, and then a 2nd rename moves the vreplicatoin table in its place. This is protected under a buffering rule on the primary, but replicas see two distinct renames and there is a point in time on the replica, where the table does not exist. Of course this means queries are failing.
There is also a scenario depicted in #11226, where even queries against the primary may fail.
This PR attempts to introduce an algorithm that is similar to the
gh-ost
cut-over. It is simplified, because the buffering rule gives us some safety net.Still work in progress as I/m seeing some flunctuations.
Related Issue(s)
#6926
Checklist
Deployment Notes