-
Notifications
You must be signed in to change notification settings - Fork 109
Conversation
@@ -164,7 +164,7 @@ type healthExecutor struct { | |||
func (p *healthExecutor) Execute(ctx context.Context) error { | |||
p.Progress.NextStep("Waiting for the planet to start") | |||
p.Info("Waiting for the planet to start.") | |||
err := utils.Retry(defaults.RetryInterval, defaults.RetryAttempts, | |||
err := utils.RetryWithInterval(ctx, defaults.ExponentialBackOff(), |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Found this issue that prevents timely abort (e.g. when explicitly aborted by a user) of the installer during the planet status wait phase.
@@ -391,113 +391,50 @@ func (s *PlanSuite) TestPlanWithIntermediateRuntimeUpdate(c *check.C) { | |||
}) | |||
} | |||
|
|||
func (s *PlanSuite) TestUpdatesEtcdFromManifestWithoutLabels(c *check.C) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Removed the test about assumed etcd version since it is expected that we have both installed and upgrade versions available.
ee273bc
to
166d95d
Compare
Otherwise, builds are subject to environment changes regarding the tele login state.
* Cherry-pick changes from #1982 to fix builds. Otherwise, builds are subject to environment changes regarding the tele login state. * Bump e
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think that this massive builder refactoring makes this a pretty risky change esp. for an LTS release. My worry is we'll break one of our subtle tele build use-cases as we did in the past.
I would either see if it's possible to fix the custom planet problem with less of unnecessary refactoring, or extensively test all tele build use-cases (for both enterprise and open-source where applicable) to make sure nothing regresses:
- That when logged out, it pulls stuff from default hub.
- That when logged into a hub (both via "tele login" and "tsh login"), it uses that hub.
- That when --hub/--token/--repository provided, it uses the indicated hub.
- Behavior of --state-dir flag, should be source for both packages and credentials (when logged in via tele login --state-dir), and when used in conjunction with --hub/--token/--repository flags.
Maybe I'm forgetting something else...
8cff16f
to
8ab796f
Compare
Checked the following for enterprise
|
8ab796f
to
cc7b7f3
Compare
Hi @a-palchikov, do you think that this pr will be a part of the next release version of |
@dorsany I have a couple more things to take care as part of this PR that were discovered in the neighbor branch, but I hope it will. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Left a couple of question, otherwise lgtm.
d84e4a8
to
b8c750c
Compare
7c19daa
to
847294c
Compare
…Func for a clearer comprehension of the builder. Expose the build API key to be able to use hub/token combo to access distribution hub regardless of the login context.
… the version file to get this detail.
Read the etcd cluster version when creating the upgrade operation to use as a fallback in case the planet image does not have the etcd version information in the package manifest as expected.
847294c
to
9bfc212
Compare
Description
Forward port of #1962.
Also fixes an issue with the upgrade unable to determine existing etcd server version if the planet container did not expose this information.
Type of change
Linked tickets and other PRs
TODOs
Testing done
Prerequisites
Clusters
6.1.33
), 3-node cluster