Description of problem:
Refer to issue https://github.com/openshift/origin/issues/8948
The logging deployment sets up ImageStreams and related DeploymentConfigs. Triggers on each DC result in deployment once the necessary tag is imported into the IS. However, it turns out that only a subset of the available tags are imported automatically (ref https://github.com/openshift/origin/issues/8952), and 3.2.0 (the current default) is frequently not among them. Thus deployment never occurs.
Version-Release number of selected component (if applicable):
Noticed most frequently with 3.2.0, however it will probably affect all versions.
More often than not. It's not clear why sometimes v3.2 is imported versus 3.2.0.
Steps to Reproduce:
1. Follow deployment instructions at https://docs.openshift.com/enterprise/3.2/install_config/aggregate_logging.html
Deployment occurs in some cases, but not others. In the cases where it doesn't, it's because the 3.2.0 tag was not imported, although "latest" and "v3.2" (which have the same image ID) are.
The quick "fix" will be to update docs to indicate the following workaround:
$ oc import-image logging-auth-proxy:3.2.0 \
$ oc import-image logging-kibana:3.2.0 \
$ oc import-image logging-elasticsearch:3.2.0 \
$ oc import-image logging-fluentd:3.2.0 \
As a more permanent fix, we will dispense with the ImageStreams and triggers, simply deploying the intended images directly.
Commit pushed to master at https://github.com/openshift/openshift-docs
logging: manually import tags for deployment
Docs fix to bug 1338965
This is an ugly workaround, but hopefully short-term.
*** Bug 1339060 has been marked as a duplicate of this bug. ***
Docs workaround at https://github.com/openshift/openshift-docs/pull/2141
FYI, The change is also needed in upgrade doc https://docs.openshift.com/enterprise/3.2/install_config/upgrading/manual_upgrades.html#manual-upgrading-efk-logging-stack
While there's a docs workaround that helps in the logging deployment (and possibly others), shouldn't this track the problem of import-image?
and if so, we should probably move it from Logging to CLI.
(In reply to Josep 'Pep' Turro Mauri from comment #5)
> While there's a docs workaround that helps in the logging deployment (and
> possibly others), shouldn't this track the problem of import-image?
> I.e. https://github.com/openshift/origin/pull/9163
I wouldn't expect that to get into Enterprise before 3.3, and by that time I fully expect to have eliminated the problem by removing the usage of ImageStreams in logging deployment.
Agree a note in the upgrade doc would help too until then.
(In reply to Xia Zhao from comment #4)
> FYI, The change is also needed in upgrade doc
Could you validate that https://github.com/openshift/openshift-docs/pull/2428 addresses the problem and the upgrade sequence works with this in place?
I think we'll want to keep this bug open to track the ultimate solution to this; docs are a stopgap.
@lmeyer The modification on upgrade doc and also the sequence mentioned looks good to me.
Note published in the upgrade docs https://docs.openshift.com/enterprise/latest/install_config/upgrading/manual_upgrades.html#manual-upgrading-efk-logging-stack
Keeping the bug open for tracking removing imagestreams entirely.
As of 3.3 imagestreams and DC triggers are not a part of the deployment, so this should not be a problem.
Yes, we never encountered this issue with the 3.3.0 logging deployer, set to verified.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.