UPSTREAM: <carry>: Don't retry storage calls with side effects. #70
Conversation
The existing patch retried any etcd error returned from storage with the code "Unavailable". Writes can only be safely retried if the client can be absolutely sure that the initial attempt ended before persisting any changes. The "Unavailable" code includes errors like "timed out" that can't be safely retried for writes.
|
/cherry-pick openshift-apiserver-4.16-kubernetes-1.29.2 |
|
@benluddy: once the present PR merges, I will cherry-pick it on top of In response to this:
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. |
|
the proof PRs are green, merging this PR. |
p0lyn0mial
merged commit f80f535
into
openshift:openshift-apiserver-4.17-kubernetes-1.29.2
|
@benluddy: new pull request created: #71 In response to this:
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. |
The existing patch retried any etcd error returned from storage with the code "Unavailable". Writes can only be safely retried if the client can be absolutely sure that the initial attempt ended before persisting any changes. The "Unavailable" code includes errors like "timed out" that can't be safely retried for writes.