-
- document creation of pull secret for the SDI Registry and linking to an OCP service account
-
- do not rely on voracluster (optional component) for ensuring pull secret
-
- when build pods are pruned manually, right after the run of observer's run script, the following may happen:
oc logs -n sdi-observer -f bc/sdi-observer Error from server (BadRequest): pods "sdi-observer-16-build" not found
-
- install jq from the regular RHEL8 repositories (rhocp-4.8-for-rhel-8-x86_64-rpms)
-
ensure clusterrolebinding to get clusteroperators/openshift-apiserver to get reliably OCP server version
-
- deploy-registry: do not redeploy if the image cannot be pulled
- if image cannot be pulled (because e.g. registry lost the blobs), an endless loop begins
-
- switch to certman
-
- modify observer's jq script for vsystem-vrep patching like this:
'. as $filtered | . +' \ '[if isempty($filtered) then {' \ '"metadata": {' \
to
'. as $filtered | . +' \ '[if isempty($filtered[]) then {' \ '"metadata": {' \
after verifying it works; as it is, it always evaluates to false
-
- determine access mode in deploy-registry script
-
- inject CA certificate into newly created tenants
-
- filter out Not found messages like the following
Error from server (NotFound): routes.route.openshift.io "vsystem" not found Error from server (NotFound): services "vsystem" not found Error from server (NotFound): secrets "ca-bundle.pem" not found Mon, 15 Mar 2021 15:05:52 +0000 Not creating vsystem route for the missing vsystem service...
-
- filter out Forbidden errors
oc logs -n sdi-observer -f dc/sdi-observer Error from server (Forbidden): clusterrolebindings.rbac.authorization.k8s.io "sdi-observer-admin-in-sdi-observer" is forbidden: User "system:serviceaccount:sdi-observer:sdi-observer" cannot get resource "clusterrolebindings" in API group "rbac.authorization.k8s.io" at the cluster scope Error from server (Forbidden): error when replacing "STDIN": clusterrolebindings.rbac.authorization.k8s.io "sdi-observer-admin-in-sdi-observer" is forbidden: User "system:serviceaccount:sdi-observer:sdi-observer" cannot update resource "clusterrolebindings" in API group "rbac.authorization.k8s.io" at the cluster scope Mon, 15 Mar 2021 19:34:27 +0000 Not replacing ClusterRole/sdi-observer-cluster-access-in-sdi-observer created by "sdi-observer-template" with a new object created by "registry-deploy". Error from server (Forbidden): clusterrolebindings.rbac.authorization.k8s.io "sdi-observer-cluster-access-in-sdi-observer" is forbidden: User "system:serviceaccount:sdi-observer:sdi-observer" cannot get resource "clusterrolebindings" in API group "rbac.authorization.k8s.io" at the cluster scope
-
- fix: Error from server (NotFound): voraclusters.sap.com "vora" not found
Names do not match (/vora != ). Something is terribly wrong!
-
- do not terminate observer when SDI or slcbridge namespaces are missing
-
- fix uninstallation
-
job datahub.checkpointstore-cleanup keeps restarting
Tue, 29 Sep 2020 16:32:17 +0000 Service account datahub-postaction-sa in sdi namespace can already pull images from sdi-observer namespace. pod "datahub.checkpointstore-cleanup-bfd3c5-9f67d8-df2sd" deleted Error from server (NotFound): jobs.batch "datahub.checkpointstore-cleanup-bfd3c5-9f67d8" not found
-
-
- delete obsolete autogenerated secrets with a command like:
# oc get secret -o json | jq -r '.items | sort_by(.metadata.creationTimestamp) | [.[] | select((.metadata.annotations["kubernetes.io/service-account.name"] // "") | test("^(sdi-observer|container-image-registry)$")) ] | group_by(.metadata.annotations["kubernetes.io/service-account.name"]) | [.[] | .[0:((. | length)-2)]] | flatten(1)[] | "\(.metadata.name): \(.metadata.creationTimestamp)"' | awk -F : '{print "secret/"$1}' | xargs -r oc delete
-
- prevent from the following build error:
2m48s Warning BuildConfigInstantiateFailed buildconfig/sdi-observer error instantiating Build from BuildConfig sdi-observer/sdi-observer (0): Error resolving ImageStreamTag ubi9:latest in namespace sdi-observer: unable to find latest tagged image
verify manual solution:
oc tag --reference-policy=local --scheduled --source=docker registry.redhat.io/ubi9/ubi:latest ubi9:latest
-
- do not re-deploy registry each time the observer is restarted
-
- add job or webhook for observer's automated updates
-
- break resource handling in observer's loop into separate modules
-
- add job for updating registry's ca bundle in image config
- make observer observe router-ca secret in openshift-ingress-operator namespace
-
- observer to grant necessary SCCs
-
- observer to granc admin role in sdi namespace to vora crd instance
-
- change RWO volumes to RWX where it makes sense