Symptoms
Application pods cannot start; facing MountVolume error after migrating from ClassicFullstack(CFS) to Cloudnative(CNF).
Error event
MountVolume.MountDevice failed for volume "pvc-xxxxxxxxxxxxxx" : rpc error: code = Internal desc = failed to get disk format of disk /dev/rbd0: blkid returns invalid output: ERROR: ld.so: object '/opt/dynatrace/oneagent/agent/bin/current/linux-x86-64/liboneagentproc.so' from /etc/ld.so.preload cannot be preloaded (cannot open shared object file): ignored. DEVNAME=/dev/rbd0 TYPE=ext4
- Error persists even after removing Dynatrace components (Namespace, DynaKube, Operator, Webhook, OneAgent and ActiveGate) and restarting application pods.
- /etc/ld.so.preload is either empty or does not exist.
- The node still has the old reference of /opt/dynatrace/oneagent/agent/ CFS, which should be changed to /opt/dynatrace/oneagent-pass/ in CNF.
Resolution
- Restart the Nodes where the application pods were deployed.
- Move forward with a clean CloudNativeFullStack Operator installation.
Note
If you are running Kubernetes clusters utilizing a CRI-O container runtime and running OneAgent version 279 or earlier, there is an additional step needed.
- Run /opt/dynatrace/oneagent/agent/uninstall.sh script on each Node, otherwise the migration fails.
See documentation
