none of the deployment works with okd 3.11 #715
Comments
|
@gacopl could you try using the manifests from the 0.8.1 release and remove all container arguments from |
|
I would have expected CI to catch this type of bug. Is that not the case? |
|
Our CI is geared towards OpenShift 4.0 and we do not re-test older releases.
…On Sat, Feb 16, 2019 at 1:02 PM ron1 ***@***.***> wrote:
I would have expected CI to catch this type of bug. Is that not the case?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#715 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AD-4LpZUmKEQHeaY_MwVOmcAp0rSqDgyks5vOEe8gaJpZM4a-sc9>
.
--
Nick Hale
|
|
@njhale ok that seemed to work at least olm is not dying now i need to figure out what's with requirementsnotmet when installing CSV |
|
@gacopl The requirement status section of the CSV's status should tell you exactly what's missing on the cluster for the CSV to run. You can either create the missing resources manually, or if you have an OLM CatalogSource that contains your CSV, you can create a Susbcription which will attempt to resolve and apply them for you. All resource generation besides APIService and Deployment are now handled by the catalog-operator and requires a Susbcription. |
|
Thanks @njhale i'm trying to wrap my head around catalogs, after installatio only pacageserver catalog is present i want to try out community operators, specificaly couchbase from certified operators, how can i add catalogs? I see the packageserver is served through grpc from some pod. In Enterprise OCP 3.11 there were special configmaps but that is more than half year old |
|
I understand that CSV is parsed and the installplan is created but this happens only when you subscribe from something from catalog, how can i add more stuff or new catalog to test out patches i made for CSV |
@njhale Thanks for the feedback. Given that OLM is in Tech Preview for OCP 3.11, is there intention to keep the latest version of OLM working on OCP 3.11 or are all efforts focused exclusively on deployments to OKD/OCP 4.0 pre-releases? |
|
@ron1 We are really just restricted in whether we depend on any backwards incompatible kubernetes changes. So we can't guarantee the latest OLM image and manifests will work with OpenShift installations based on older kubernetes versions. Our previous release manifests are tied to specific OLM image digests, so if a version of these manifests works on 3.11 it should continue to work, unless the manifests were changed somewhere along the way (need to double check this hasn't already happened). |
@gacopl If you just want to try out community-operators with a newer version of OLM apiVersion: operators.coreos.com/v1alpha1
kind: CatalogSource
metadata:
name: community-operators
spec:
displayName: Community Operators
image: quay.io/njhale/community-operators@sha256:37f1dd6ab4f1082af9d8f9ef028a2be4fb2837c5a75ba59bd127ebc723bfee8d
publisher: community-operators
sourceType: grpcWhen you create a new subscription to the operator you want, be sure to specify the correct |
|
@njhale Does it make sense that 0000_50_olm_14-operatorstatus.yaml fails to apply on OCP 3.11? I see that this file exists for OCP but not OKD. Am I correct this file exclusively targets OCP 4.0? Also would you mind describing the process you used to create the community-operators image referenced above? Am I correct you used a variation of https://github.com/operator-framework/operator-registry/blob/master/upstream.Dockerfile with some additional steps? Finally, when I created the community-operators CatalogSource you provided above, all my packageserver pods immediately started panicking with the following stack trace. Any thoughts? |
|
@ron1 your first point is correct. You are also correct in that https://github.com/operator-framework/operator-registry/blob/master/upstream.Dockerfile is the "basis" for how I packaged community-operators as an OLM catalog - it provides an example of how to build a an OLM operator-registry image, which is OLM's preferred way to package operator catalog content. We have a PR in-flight that should be merging soon to update the docs in that repo to better reflect this. As for the 3rd issue - it seems like the version of OLM being used is older than what's currently in master. From the provided panic log, registry.go:166 is supposed to be accessing a nil pointer, but in master this seems to be a log call. We also have a check early in this function to bail out early if the |
|
@gacopl I just tested the apiVersion: operators.coreos.com/v1alpha1
kind: Subscription
metadata:
name: etcd
namespace: openshift-operator-lifecycle-manager
spec:
source: rh-operators
sourceNamespace: openshift-operator-lifecycle-manager
name: etcd
channel: alphaHow did you install OKD? At this point, I'm reasonably sure OLM's manifests and images are good for |
|
basic oc cluster up of 3.11, the 0,8x branch worked for me after doing the args fix |
|
@njhale Does OLM 0.7.4 support the latest CSV schemas used by operators currently in the community-operators repository? Also, does it support the operator-registry-based CatalogSources with sourceType grpc you described in your prior comment? |
|
|
@njhale Given that OLM has changed significantly between 0.7.4 and 0.8.1+ including grpc CatalogSources, OperatorGroups, InstallModes, etc., and given that Operators currently in the community-operators repo are likely being tested only against OLM 0.8.1+, would you expect OLM 0.7.4 to reliably manage the current set of Operators in the community-operators repo? If so, what is the best way to assemble/deploy community-operators/upstream-community-operators into a ConfigMap-based CatalogSource for use by OLM 0.7.4? |
|
Current status for 3.11:
If you need to play with OLM and are okay with those caveats on 3.11, you might try the upstream installation instructions: https://github.com/operator-framework/operator-lifecycle-manager/releases/tag/0.10.0 |
latest deployment for okd makes packageservice restarts
0.7.4 deployment make operator crashloopbackoff
catalogs don't work, and issuing CSV alone to openshift-operators fails, in cluster console failed staus and none deployments are created
any ideas how to install OLM with okd 3.11?
The text was updated successfully, but these errors were encountered: