Symptoms
Sometimes, when vCMP guests are started after the vCMP host has been upgraded, the guests may enter an unhealthy state due to clusterd constantly restarting.
Impact
-- This issue might prevent the guest from servicing traffic if
both slots fail to become operable because the system failed to load the config and clusterd keeps restarting.
-- During boot up of the vCMP guest, the following message is logged to /var/log/ltm:
err mcpd[6519]: 0107146f:3: Self-device state mirroring address cannot reference the non-existent Self IP ([IP address]); Create it in the /Common folder first.
-- /var/log/ltm shows clusterd constantly restarting.
-- vCMP guest may show one or more slots in INOPERATIVE state while vCMP host shows slots as RUN/Healthy.
Conditions
-- vCMP guest has Mirroring IP configured.
-- vCMP guest is powered on after vCMP host upgrade.
Workaround
-- To avoid the issue before it occurs:
1. Prior to shutting down vCMP guests before vCMP host upgrade, ensure guests have free space in the /var partition.
2. Issue 'tmsh save sys config' on the vCMP guest.
3. Issue 'ls /var/db/mcp*' and confirm the presence of mcpdb.bin and mcpdb.info in the /var/db directory.
4. Proceed with vCMP guest shutdown and vCMP host upgrade as per standard F5 recommended process.
-- To mitigate after the issue has been experienced on a vCMP guest:
1. Set the vCMP guest to the Configured state and wait for it to complete transition to Configured.
2. Set vCMP guest to Deployed state.
3. Review boot up log and confirm 'Self-device state mirroring address cannot reference the non-existent Self IP' message is no longer present.
4. Review /var/log/ltm and confirm clusterd is no longer restarting.
5. If issue persists, delete and recreate the vCMP guest
