UPSTREAM: <drop>: Double the global timeout if performing a global list #15505
Conversation
openshift-merge-robot
added
the
size/XS
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: smarterclayton No associated issue. Update pull-request body to add a reference to an issue, or get approval with The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these OWNERS Files:
You can indicate your approval by writing |
|
|
||
| // when performing global calls, increase the timeout for very large clusters | ||
| timeout := globalTimeout | ||
| if requestInfo.IsResourceRequest && requestInfo && len(requestInfo.Namespace) == 0 { |
There was a problem hiding this comment.
intent is good, but this doesn't compile
openshift-merge-robot
added
the
approved
| if requestInfo.IsResourceRequest && len(requestInfo.Namespace) == 0 { | ||
| timeout = timeout * 2 | ||
| } | ||
|
|
||
| return time.After(globalTimeout), apierrors.NewServerTimeout(schema.GroupResource{Group: requestInfo.APIGroup, Resource: requestInfo.Resource}, requestInfo.Verb, 0) |
There was a problem hiding this comment.
Use the local var here, right?
In 3.6, users still on etcd may be seeing large list calls take in excess of 1 minute (the hard limit). In 3.7, all users will be on etcd3 and protobuf and we expect times to drop by an order of magnitude. Since our largest in the wild clusters are taking about 1 minute 3 seconds, double the global timeout limit for global lists only for 3.6, and in 3.7 this will no longer be necessary.
|
journald log snip /retest |
|
/retest |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: smarterclayton No associated issue. Update pull-request body to add a reference to an issue, or get approval with The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these OWNERS Files:
You can indicate your approval by writing |
openshift-merge-robot
removed
the
approved
In 3.6, users still on etcd may be seeing large list calls take in
excess of 1 minute (the hard limit). In 3.7, all users will be on etcd3
and protobuf and we expect times to drop by an order of magnitude.
Since our largest in the wild clusters are taking about 1 minute 3
seconds, double the global timeout limit for global lists only for 3.6,
and in 3.7 this will no longer be necessary.
@deads2k