Some hints and guides for when you got stuck during deployment and daily use of AWX.
- Troubles during Deployment
- First Step: Investigate your Situation
- Reveal "censored" output in the AWX Operator's log
- The Pod is
ErrImagePullwith "429 Too Many Requests" - The Pod is
Pendingwith "1 Insufficient cpu, 1 Insufficient memory." event - The Pod is
Pendingwith "1 pod has unbound immediate PersistentVolumeClaims." event - The Pod is
Runningbut stucked with "wait-for-migrations" message - The Pod for PostgreSQL is in
CrashLoopBackOffstate and shows "Permission denied" log
- Troubles during Daily Use
You can start investigating troubles during deployment with following two things.
- Status and Events of the Pods
- Logs of the Containers inside the Pods
First, check the STATUS for the Pods by this command.
kubectl -n awx get podIf the Pods are working properly, its STATUS are Running. If your Pods are not in Running state e.g. Pending, ImagePullBackOff or CrashLoopBackOff etc., the Pods might have some problems. In the following example, the Pod awx-84d5c45999-h7xm4 is in Pending state.
$ kubectl -n awx get pod
NAME READY STATUS RESTARTS AGE
awx-operator-controller-manager-57867569c4-ggl29 2/2 Running 0 8m20s
awx-postgres-15-0 1/1 Running 0 7m26s
awx-task-5d8cd9b6b9-8ptjt 0/4 Pending 0 6m55s
awx-web-66f89bc9cf-6zck5 0/3 Pending 0 6m9sIf you have the Pods which has the unexpected state instead of Running, the next step is checking Events for the Pod. The command to get Events for the pod is:
kubectl -n awx describe pod <Pod Name>By this command, you can get the Events for the Pod you specified at the end of the output.
$ kubectl -n awx describe pod awx-task-5d8cd9b6b9-8ptjt
...
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Warning FailedScheduling 106s default-scheduler 0/1 nodes are available: 1 Insufficient cpu, 1 Insufficient memory. 👈👈👈
Warning FailedScheduling 105s default-scheduler 0/1 nodes are available: 1 Insufficient cpu, 1 Insufficient memory. 👈👈👈In most cases, you can find the reason why the Pod is not Running from Events. In the example above, I can see that it is due to lack of CPU or memory.
The logs also helpful to get the reason why something went wrong. In particular, if the status of the Pod is Running but the Pod does not works as expected, you should check the logs.
The commands to get the logs are following. -f is optional, useful to watch the logs as well as get the logs.
# Get the logs of specific Pod.
# If the Pod includes multiple containers, container name has to be specified.
kubectl -n awx logs -f <POD>
kubectl -n awx logs -f <POD> -c <CONTAINER>
# Get the logs of specific Pod which is handled by Deployment resource.
# If the Pod includes multiple containers, container name has to be specified.
kubectl -n awx logs -f deployment/<DEPLOYMENT>
kubectl -n awx logs -f deployment/<DEPLOYMENT> -c <CONTAINER>
# Get the logs of specific Pod which is handled by StatefulSet resource
# If the Pod includes multiple containers, container name has to be specified.
kubectl -n awx logs -f statefulset/<STATEFULSET>
kubectl -n awx logs -f statefulset/<STATEFULSET> -c <CONTAINER>For AWX Operator and AWX, specifically, the following commands are helpful.
- Logs of AWX Operator
kubectl -n awx logs -f deployment/awx-operator-controller-manager
- Logs of AWX related init containers
kubectl -n awx logs -f deployment/awx-task -c initkubectl -n awx logs -f deployment/awx-task -c init-projects
- Logs of AWX related containers
kubectl -n awx logs -f deployment/awx-web -c awx-webkubectl -n awx logs -f deployment/awx-web -c awx-rsyslogkubectl -n awx logs -f deployment/awx-web -c rediskubectl -n awx logs -f deployment/awx-task -c awx-taskkubectl -n awx logs -f deployment/awx-task -c awx-eekubectl -n awx logs -f deployment/awx-task -c awx-rsyslogkubectl -n awx logs -f deployment/awx-task -c redis
- Logs of PostgreSQL
kubectl -n awx logs -f statefulset/awx-postgres-15
If you've found the FAILED tasks while investigating AWX Operator's log, sadly sometimes it's marked as censored and you can't get actual log.
$ kubectl -n awx logs -f deployments/awx-operator-controller-manager
...
TASK [Restore database dump to the new postgresql container] ********************************
fatal: [localhost]: FAILED! => {"censored": "the output has been hidden due to the fact that 'no_log: true' was specified for this result", "changed": true}
...AWX Operator 0.23.0 or later supports making this revealed.
To achieve this, you can uncomment no_log: false manually under spec for your awx.yaml, awxbackup.yaml, or awxrestore.yaml, and then re-run your deployment, backup, or restoration.
...
spec:
...
# Uncomment to reveal "censored" logs
no_log: false 👈👈👈
...If your Pod for PostgreSQL is in ErrImagePull and its Events shows following events, this is due to the rate limit on Docker Hub.
$ kubectl -n awx describe pod awx-postgres-13-0
...
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Pulling 9s kubelet Pulling image "postgres:13"
Warning Failed 2s kubelet Failed to pull image "postgres:13": rpc error: code = Unknown desc = failed to pull and unpack image "docker.io/library/postgres:13": failed to copy: httpReadSeeker: failed open: unexpected status code https://registry-1.docker.io/v2/library/postgres/manifests/sha256:...: 429 Too Many Requests - Server message: toomanyrequests: You have reached your pull rate limit. You may increase the limit by authenticating and upgrading: https://www.docker.com/increase-rate-limit
Warning Failed 2s kubelet Error: ErrImagePull
Normal BackOff 1s kubelet Back-off pulling image "postgres:13"
Warning Failed 1s kubelet Error: ImagePullBackOffIf you follow the steps in this repository to deploy you AWX, your pull request to Docker Hub will be identified as a free, anonymous account. Therefore, you will be limited to 200 requests in 6 hours. The message "429 Too Many Requests" indicates that it has been exceeded.
To solve this, you can simply wait until the limit is freed up, or consider giving your Docker Hub credentials to K3s by following the guide on this page.
If your Pod is in Pending state and its Events shows following events, the reason is that the node does not have enough CPU and memory to start the Pod. By default AWX requires at least 2 CPUs and 4 GB RAM. In addition more resources are required to run K3s and the OS itself.
$ kubectl -n awx describe pod awx-task-5d8cd9b6b9-8ptjt
...
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Warning FailedScheduling 106s default-scheduler 0/1 nodes are available: 1 Insufficient cpu, 1 Insufficient memory. 👈👈👈
Warning FailedScheduling 105s default-scheduler 0/1 nodes are available: 1 Insufficient cpu, 1 Insufficient memory. 👈👈👈Typical solutions are one of the following:
- Add more CPUs or memory to your K3s node.
- If you have at least 3 CPUs and 5 GB RAM, AWX may work.
- Reduce resource requests for the containers.
-
The minimum resource requirements can be ignored by adding three lines in
base/awx.yml.... spec: ... web_resource_requirements: {} 👈👈👈 task_resource_requirements: {} 👈👈👈 ee_resource_requirements: {} 👈👈👈 init_container_resource_requirements: {} 👈👈👈 postgres_init_container_resource_requirements: {} 👈👈👈 postgres_resource_requirements: {} 👈👈👈 redis_resource_requirements: {} 👈👈👈 rsyslog_resource_requirements: {} 👈👈👈
-
You can specify more specific value for each containers. Refer official documentation for details.
-
In this way you can run AWX with fewer resources, but you may encounter performance issues.
-
If your Pod is in Pending state and its Events shows following events, the reason is that no usable Persistent Volumes are available.
$ kubectl -n awx describe pod awx-task-5d8cd9b6b9-8ptjt
...
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Warning FailedScheduling 24s default-scheduler 0/1 nodes are available: 1 pod has unbound immediate PersistentVolumeClaims. 👈👈👈Check the STATUS of your PVs and ensure your PVs doesn't have Available or Bound state.
$ kubectl get pv
NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE
awx-projects-volume 2Gi RWO Retain Released awx/awx-projects-claim awx-projects-volume 17h
awx-postgres-15-volume 8Gi RWO Retain Released awx/postgres-15-awx-postgres-15-0 awx-postgres-volume 17hProbably this is the second (or more) time to deploy AWX for you. These PVs which have Released state are tied to your old (and probably no longer exists now) PVCs you created in the past.
There are a few things you should to know about the PVs in Kubernetes.
- Once a PV is bound from a PVC, it keeps the PVC name in its
claimRefentry. This will be shown in theCLAIMcolumn in the result of the commandkubectl get pv. - The
Releasedstate of the PV means that the PV was bound by PVC in theclaimRefentry in the past but now the PVC does not exist. The PV in this state cannot be bound by any PVC other than the one recorded inclaimRef. - To allow the PV to bind from a PVC other than the one recorded in
claimRef, theclaimRefentry must be empty and the PV must hasAvailablestate.
To solve this, typical solutions are one of the following:
-
Patch the PV to empty
claimRefentry for the PV.-
Invoke following commands:
kubectl patch pv <PV Name> -p '{"spec":{"claimRef": null}}'
-
-
Delete the PV and recreate it.
-
Invoke following commands:
# Delete the PV kubectl delete pv <PV Name> # Recreate the PV kubectl apply -k base
-
Sometimes your AWX pod is Running state correctly but not functional at all, and its log shows following message repeatedly.
kubectl -n awx logs -f deployment/awx-web -c awx-web
[wait-for-migrations] Waiting for database migrations...
[wait-for-migrations] Attempt 1 of 30
[wait-for-migrations] Waiting 0.5 seconds before next attempt
[wait-for-migrations] Attempt 2 of 30
[wait-for-migrations] Waiting 1 seconds before next attempt
[wait-for-migrations] Attempt 3 of 30
[wait-for-migrations] Waiting 2 seconds before next attempt
[wait-for-migrations] Attempt 4 of 30
[wait-for-migrations] Waiting 4 seconds before next attempt
...
[wait-for-migrations] Attempt 28 of 30
[wait-for-migrations] Waiting 30 seconds before next attempt
[wait-for-migrations] Attempt 29 of 30
[wait-for-migrations] Waiting 30 seconds before next attempt
...This problem occurs when the AWX pod and the PostgreSQL pod cannot communicate properly. In most cases, the cause of this is the network on your K3s.
To solve this, check or try the following:
- Ensure your PostgreSQL (typically the Pod named
awx-postgres-0,awx-postgres-13-0, orawx-postgres-15-0) is inRunningstate. - Ensure
hostunderawx-postgres-configurationinbase/kustomization.yamlhas correct value.- Specify
awx-postgresfor AWX Operator 0.25.0 or earlier,awx-postgres-13for0.26.0to2.12.2,awx-postgres-15for newer versions.
- Specify
- Ensure your
firewalld,ufwor any kind of firewall has been disabled on your K3s host. - Ensure your
nm-cloud-setupservice on your K3s host is disabled if exists. - Ensure your
awx-postgres-configurationhas correct values, especially if you're using external PostgreSQL. - Uninstall K3s and install it again.
In this situation, your Pod for PostgreSQL is in CrashLoopBackOff state and its log shows following error message.
$ kubectl -n awx get pod
NAME READY STATUS RESTARTS AGE
awx-operator-controller-manager-57867569c4-ggl29 2/2 Running 0 8m20s
awx-postgres-13-0 1/1 CrashLoopBackOff 5 7m26s
awx-task-5d8cd9b6b9-8ptjt 0/4 Running 0 6m55s
awx-web-66f89bc9cf-6zck5 0/3 Running 0 6m9s
# On PostgreSQL 13
$ kubectl -n awx logs statefulset/awx-postgres-13
mkdir: cannot create directory '/var/lib/postgresql/data': Permission denied
# On PostgreSQL 15
$ kubectl -n awx logs statefulset/awx-postgres-15
mkdir: cannot create directory '/var/lib/pgsql/data/userdata': Permission deniedYou should check the permissions and the owner of directories where used as PV on your K3s host.
For the PostgreSQL 13 that deployed by AWX Operator 2.12.2 or earlier, if you followed my guide, it would be /data/postgres-13. There is additional data directory created by K3s under /data/postgres-13.
$ ls -ld /data/postgres-13 /data/postgres-13/data
drwxr-xr-x. 2 root root 18 Aug 20 10:09 /data/postgres-13
drwxr-xr-x. 3 root root 20 Aug 20 10:09 /data/postgres-13/dataFor example, 755 and root:root (0:0) should work. So you can try following commands.
sudo chown 0:0 /data/postgres-13 /data/postgres-13/data
sudo chmod 755 /data/postgres-13 /data/postgres-13/dataOr, you can also try 999:0 as owner/group for the directory. 999 is the UID of the postgres user which used in the container.
sudo chown 999:0 /data/postgres-13 /data/postgres-13/data
sudo chmod 755 /data/postgres-13 /data/postgres-13/dataFor the PostgreSQL 15 that deployed by AWX Operator 2.13.0 or later, if you followed my guide, it would be /data/postgres-15. There is additional data directory created by K3s under /data/postgres-15.
$ ls -ld /data/postgres-15 /data/postgres-15/data
drwxr-xr-x. 2 root root 18 Aug 20 10:09 /data/postgres-15
drwxr-xr-x. 3 26 root 20 Aug 20 10:09 /data/postgres-15/dataFor example, 700 and 26:0 should work. So you can try following commands. 26 is the UID of the user which used in the container.
sudo chown 26:0 /data/postgres-15 /data/postgres-15/data
sudo chmod 700 /data/postgres-15 /data/postgres-15/dataIf the job is invoked to a large number of hosts or is running long time, sometimes the job is marked as failed and no log will be displayed in the Output tab.
This is a problem caused by log rotation on Kubernetes. Refer ansible/awx#10366 for details.
In the case of K3s, you can reduce the possibility of this issue by changing the configuration as follows.
# Change configuration using script:
$ curl -sfL https://get.k3s.io | sh -s - --write-kubeconfig-mode 644 --kubelet-arg "container-log-max-files=4" --kubelet-arg "container-log-max-size=50Mi"
# If you don't want to use the script, modify /etc/systemd/system/k3s.service manually:
$ cat /etc/systemd/system/k3s.service
...
ExecStart=/usr/local/bin/k3s \
server \
'--write-kubeconfig-mode' \
'644' \
'--kubelet-arg' \ 👈👈👈
'container-log-max-files=4' \ 👈👈👈
'--kubelet-arg' \ 👈👈👈
'container-log-max-size=50Mi' \ 👈👈👈Then restart K3s. The K3s service can be safely restarted without affecting the running resources.
sudo systemctl daemon-reload
sudo systemctl restart k3sIf you use Traefik which is K3s' Ingress controller as completely default, the Pod may not be able to get the client's IP address (see k3s-io/k3s#2997 for details). Therefore, the feature called Provisioning Callback in AWX does not work properly since AWX can't determine actual IP address of the remote host who request callback.
For this reason, you should fix the Traefik configuration. For a single node like doing in this repository, reconfiguring your Traefik by creating YAML file is the easy way.
sudo tee /var/lib/rancher/k3s/server/manifests/traefik-config.yaml <<EOF
---
apiVersion: helm.cattle.io/v1
kind: HelmChartConfig
metadata:
name: traefik
namespace: kube-system
spec:
valuesContent: |-
hostNetwork: true
EOFThen wait until your traefik by the following command is 1/1 READY.
kubectl -n kube-system get deployment traefikNow your client's IP address can be passed correctly through X-Forwarded-For and X-Real-Ip headers.
The last step is modifying AWX. By default, AWX uses only REMOTE_ADDR and REMOTE_HOST headers to determine the remote host (means HTTP client). Therefore, you have to make AWX to use X-Forwarded-For header.
This can be achieved by modifying your base/awx.yaml and apply it, or simply configure Remote Host Headers via AWX UI.
If you want to use base/awx.yaml to achieve this, add following three lines to your base/awx.yaml.
...
spec:
...
extra_settings: 👈👈👈
- setting: REMOTE_HOST_HEADERS 👈👈👈
value: "['HTTP_X_FORWARDED_FOR', 'REMOTE_ADDR', 'REMOTE_HOST']" 👈👈👈Then apply this change and wait for your AWX will be reconfigured.
kubectl apply -k baseYou can watch its progress by following command as did when you deploy AWX at the first time.
kubectl -n awx logs -f deployments/awx-operator-controller-managerAlternatively you can modify this settings via AWX UI. Move on to Settings > Miscellaneous System settings > Edit page, then and put following JSON strings as Remote Host Headers.
[
"HTTP_X_FORWARDED_FOR",
"REMOTE_ADDR",
"REMOTE_HOST"
]Now your Provisioning Callback should work. In my environment, the name of the host in the inventory have to be defined using IP address instead of DNS hostname.
The job failed and I got "ERROR! couldn't resolve module/action" or "Failed to import the required Python library" message
When you launch the Job Template, it may fail and you will see an error like the following:
ERROR! couldn't resolve module/action 'community.postgresql.postgresql_info'. This often indicates a misspelling, missing collection, or incorrect module path.
The error appears to be in '/runner/project/site.yml': line 6, column 7, but may
be elsewhere in the file depending on the exact syntax problem.
The offending line appears to be:
tasks:
- community.postgresql.postgresql_info:
^ here
Alternatively, the import of Python modules may fail.
...
TASK [community.postgresql.postgresql_info] ************************************
fatal: [localhost]: FAILED! => {"changed": false, "msg": "Failed to import the required Python library (psycopg2) on automation-job-12-v2gvf's Python /usr/bin/python3. Please read the module documentation and install it in the appropriate location. If the required library is installed, but Ansible is using the wrong Python interpreter, please consult the documentation on ansible_python_interpreter"}
...
To solve this, you can build your own Execution Image, or place requirements.yml on the specific path in your project. Refer to the guide about Execution Environment on this repository for details.