Hide Forgot
Created attachment 1227996 [details] ansbile task logs Description of problem: Incorrect pvc name was generated when deploy logging with ansible, it should be "logging-es-1" instead of "logging-es1". Issue did not repro when manually deploy logging with "--from-literal es-pvc-dynamic=true --from-literal es-pvc-size=10G" specified in deployer configmap. Logging got deployed with a wrong pvc name: # oc get po NAME READY STATUS RESTARTS AGE logging-curator-1-l0fzh 1/1 Running 0 4h logging-deployer-dlnow 0/1 Completed 0 4h logging-es-1yl4qhyc-1-91pkd 1/1 Running 0 4h logging-fluentd-5f3s1 1/1 Running 0 4h logging-fluentd-dhojf 1/1 Running 0 4h logging-kibana-1-tz7ks 2/2 Running 0 4h Version-Release number of selected component (if applicable): # openshift version openshift v3.4.0.32+d349492 kubernetes v1.4.0+776c994 etcd 3.1.0-rc.0 openshift-ansible-3.4.31-1.git.0.4b78e57 How reproducible: Always Steps to Reproduce: 1. Deploy logging with these parameters passing to ansible: openshift_hosted_logging_deployer_prefix: ${ops-registry-image-prefix} openshift_hosted_logging_deploy: true openshift_hosted_logging_storage_kind: dynamic openshift_hosted_logging_deployer_version: 3.4.0 2. Check logging pods: # oc get po NAME READY STATUS RESTARTS AGE logging-curator-1-l0fzh 1/1 Running 0 4h logging-deployer-dlnow 0/1 Completed 0 4h logging-es-1yl4qhyc-1-91pkd 1/1 Running 0 4h logging-fluentd-5f3s1 1/1 Running 0 4h logging-fluentd-dhojf 1/1 Running 0 4h logging-kibana-1-tz7ks 2/2 Running 0 4h 3. Check deployer configmap content # oc get configmap logging-deployer -o yaml apiVersion: v1 data: es-cluster-size: "1" es-ops-cluster-size: "1" es-ops-pvc-dynamic: "true" es-ops-pvc-prefix: logging-es es-ops-pvc-size: 10Gi es-pvc-dynamic: "true" es-pvc-prefix: logging-es es-pvc-size: 10Gi kibana-hostname: kibana.{subdomain} kibana-ops-hostname: kibana-ops.{subdomain} public-master-url: https://{master-domain}:8443 kind: ConfigMap metadata: creationTimestamp: 2016-12-05T02:12:33Z name: logging-deployer namespace: logging resourceVersion: "1221" selfLink: /api/v1/namespaces/logging/configmaps/logging-deployer uid: 4865c5ef-ba90-11e6-9243-42010af00014 4. Check the pvc name Actual results: 4. pvc name is logging-es1 # oc get pvc NAME STATUS VOLUME CAPACITY ACCESSMODES AGE logging-es1 Bound pvc-66fbe5b6-ba90-11e6-9243-42010af00014 10Gi RWO 4h Expected results: 4. pvc name should be logging-es-1 Additional info: 1. ansible log attached 2. deployer pod log attached
Created attachment 1227997 [details] deployer log
Per the documentation of the EFK deployer, this is working as expected. The prefix specified is followed by a number. To see the pvc name as 'logging-es-1' the prefix should be 'logging-es-'.
So the thing is, during logging deployment, if I didn't explicitly specify a value to es-ops-pvc-prefix, just specify this: --from-literal es-pvc-dynamic=true --from-literal es-pvc-size=1G deployer will create the pvc with value "logging-es-" (which is expected) but ansible created it with "logging-es", which omitted the "-" in the end. The behaviour is different here.