I’m currently testing awxbackup and awxrestore. During the awxrestore process, the awx-task and awx-web deployments are set to replicas=0, causing the pods to terminate. However, while the database is being restored, these pods unexpectedly restart and enter a running state before the restoration is complete. Consequently, the awxrestore process doesn’t finish and seems to loop indefinitely. Could this premature running state of the web and task pods be causing the restoration to fail?
This behavior has been observed by others in the community. For instance, an issue was reported where the AWX task and web pods incorrectly scaled up during the restore process, leading to interference with the database restoration.
To address this, a recommended approach is to set the AWX Custom Resource Definition (CRD) replicas value to 0 until the database restoration is fully completed. This ensures that the pods do not attempt to connect to the database prematurely. Once the restoration process concludes, you can set the replicas value back to its original state.
Implementing this strategy should help in preventing the web and task pods from starting before the database is fully restored, thereby avoiding potential restoration failures.