Description of problem: When deploying the ansible service broker, it is deployed on the compute nodes. There is no way to force it to install on, say, an type=infra node. Version-Release number of the following components: rpm -q openshift-ansible 3.9.7 rpm -q ansible 2.4.3 How reproducible: Steps to Reproduce: 1. Use o-a to deploy ansible service broker 2. oc adm manage-nodes --list-pods --selector=type=compute Actual results: Please include the entire output from the last TASK line through the end of output if an error is generated [root@ded-int-gcp-master-esjli ~]# oc adm manage-node --list-pods --selector=type=compute Listing matched pods on node: ded-int-gcp-node-compute-35rfm NAMESPACE NAME READY STATUS RESTARTS AGE logging logging-fluentd-sq9xn 1/1 Running 0 23h openshift-ansible-service-broker asb-etcd-1-kggj9 1/1 Running 0 23h Listing matched pods on node: ded-int-gcp-node-compute-dl7vw NAMESPACE NAME READY STATUS RESTARTS AGE logging logging-fluentd-ff5fn 1/1 Running 0 23h Listing matched pods on node: ded-int-gcp-node-compute-h4ssk NAMESPACE NAME READY STATUS RESTARTS AGE logging logging-fluentd-xnpwn 1/1 Running 0 23h Listing matched pods on node: ded-int-gcp-node-compute-v9dzh NAMESPACE NAME READY STATUS RESTARTS AGE logging logging-fluentd-4gtfv 1/1 Running 0 23h openshift-ansible-service-broker asb-1-deploy 0/1 Error 0 23h Expected results: No asb pods on compute nodes! Additional info: Please attach logs from ansible-playbook with the -vvv flag
https://github.com/openshift/openshift-ansible/pull/7575
Commits pushed to master at https://github.com/openshift/openshift-ansible https://github.com/openshift/openshift-ansible/commit/8b9250d37b4d7f8977ade6ba719c94516a53ea14 Bug 1557516- ASB now scheduled on infra nodes https://github.com/openshift/openshift-ansible/commit/82887b1fbe475a550b3bde12d17aea1dc38afbee Merge pull request #7575 from fabianvf/bz1557516 Bug 1557516- ASB now scheduled on infra nodes
https://github.com/openshift/openshift-ansible/pull/7692 release-3.9 backport
Retest this bug with openshift-ansible-3.9.24-1.git.0.d0289ea.el7.noarch, FAIL. According the PR, without ansible_service_broker_node_selector setting, asb pod should be landed onto region=infra node by default, with ansible_service_broker_node_selector setting, asb pod should be landed onto specified node. But in my testing, whatever set ansible_service_broker_node_selector={"role": "node"} or not, asb pod always land onto compute nodes. # oc describe po/asb-1-k992z -n openshift-ansible-service-broker <--snip--> Node-Selectors: node-role.kubernetes.io/compute=true <--snip--> Feel like the PR never being merged, but I checked my running installer, the PR is already there.
Can you post your inventory for the failed run?
Created attachment 1425570 [details] installation log with inventory embeded
it looks like you have ansible_service_broker_node_selector={"role": "node"} set in the inventory. If you don't set ansible_service_broker_node_selector at all, I think you will get the behavior you are expecting. Does this work for you?
Never mind, I was confused, I think I see the issue. The node selector might need to be specified on the podspec rather than the dc spec.
https://github.com/openshift/openshift-ansible/pull/8117
So far, the installer rpm version is openshift-ansible-3.9.27-1.git.0.52e35b5.el7.noarch in latest puddle, waiting for newer puddle.
Verified this bug with openshift-ansible-3.9.28-1.git.0.4fc2ce4.el7.noarch, and PASS. Because images for v3.9.28 is not built or unavailable on aws-reg registry, so use openshift-ansible-3.9.28-1.git.0.4fc2ce4.el7.noarch installer + 3.9/v3.9.27-1_2018-04-26.2 puddle for installation. Scenarios 1: 1. don't set ansible_service_broker_node_selector at all in inventory file, trigger installation. 2. after installation, checking: [root@qe-jialiu392-master-etcd-1 ~]# oc get po -n openshift-ansible-service-broker NAME READY STATUS RESTARTS AGE asb-1-deploy 0/1 Pending 0 3m asb-etcd-1-deploy 0/1 Pending 0 3m # oc describe po asb-etcd-1-deploy -n openshift-ansible-service-broker Name: asb-etcd-1-deploy <--snip--> Node-Selectors: region=infra Tolerations: <none> Events: Type Reason Age From Message ---- ------ ---- ---- ------- Warning FailedScheduling 2s (x18 over 4m) default-scheduler 0/2 nodes are available: 2 CheckServiceAffinity, 2 MatchNodeSelector. # oc get node -l region=infra No resources found. # oc get node NAME STATUS ROLES AGE VERSION qe-jialiu392-master-etcd-1 Ready master 17m v1.9.1+a0ce1bc657 qe-jialiu392-node-registry-router-1 Ready compute 17m v1.9.1+a0ce1bc657 The default "region=infra" node selector take effect, the "pending" behavior is expected. Scenario 2: 1. set ansible_service_broker_node_selector={"role": "node"} in inventory file, trigger installation. 2. after installation, checking: # oc get po -n openshift-ansible-service-broker NAME READY STATUS RESTARTS AGE asb-1-rftf6 1/1 Running 1 58m asb-etcd-1-vzlk6 1/1 Running 0 58m # oc describe po asb-1-rftf6 -n openshift-ansible-service-broker Name: asb-1-rftf6 Namespace: openshift-ansible-service-broker Node: qe-jialiu391-node-registry-router-1/10.240.0.22 <--snip--> Node-Selectors: role=node <--snip--> # oc get node -l role=node NAME STATUS ROLES AGE VERSION qe-jialiu391-master-etcd-1 Ready master 1h v1.9.1+a0ce1bc657 qe-jialiu391-node-registry-router-1 Ready compute 1h v1.9.1+a0ce1bc657 Based on currently only openshift-ansible-3.9.27 is attached to 33431 advisory, I move this bug to "MODIFIED", once 3.9.28 build is attached, will verify this bug.
openshift-ansible build is not attached to the advisory yet, once attached, will re-run testing.
Verified this bug with openshift-ansible-3.9.40-1.git.0.188c954.el7.noarch, and PASS. Scenario 2: 1. set ansible_service_broker_node_selector={"role": "node"} in inventory file, trigger installation. 2. after installation, checking: # oc get po -n openshift-ansible-service-broker NAME READY STATUS RESTARTS AGE asb-1-p25g2 1/1 Running 0 18h asb-etcd-1-54sv4 1/1 Running 0 18h # oc describe po asb-1-p25g2 -n openshift-ansible-service-broker <--snip--> Node-Selectors: role=node <--snip--> Base on my verification and comment 18, move this bug 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. https://access.redhat.com/errata/RHBA-2018:2335