-
Notifications
You must be signed in to change notification settings - Fork 4.1k
changefeedccl: make clock skew errors non-permanent #146754
Copy link
Copy link
Open
Labels
A-cdcChange Data CaptureChange Data CaptureC-bugCode not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior.Code not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior.E-quick-winLikely to be a quick win for someone experienced.Likely to be a quick win for someone experienced.T-cdcbranch-masterFailures and bugs on the master branch.Failures and bugs on the master branch.
Description
Metadata
Metadata
Assignees
Labels
A-cdcChange Data CaptureChange Data CaptureC-bugCode not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior.Code not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior.E-quick-winLikely to be a quick win for someone experienced.Likely to be a quick win for someone experienced.T-cdcbranch-masterFailures and bugs on the master branch.Failures and bugs on the master branch.
Currently if changefeeds encounter clock skew errors such as (
select-job: remote wall time is too far ahead (251.725287ms) to be trustworthy), they will treat it as non-transient and fail themselves. This type of error (possibly all rpc errors?) should be treated as transient so the feed can eventually recover when the cluster recovers.ref: https://cockroachlabs.slack.com/archives/C2C5FKPPB/p1747236668939499?thread_ts=1747236583.901299&cid=C2C5FKPPB
Jira issue: CRDB-50689