onemonitord in-memory db possible issues with HA

[tinova]

Description

When using HA there are some situations that lead to certain hosts containing duplicated monitord client processes with old host_ids no longer present in the host_pool database.

To Reproduce

In HA setup, remove a host from OpenNebula. In some situations this leads to duplicate "monitord" processes running on several hypervisors. There are two processes on each hypervisor: one with the current host ID and the other with the old host ID. The old host ID indicates that these machines were previously part of the zone but were removed for maintenance.

The onemonitord has in memory DB. It should be updated immediately after start from oned.

Expected behavior

Don't monitor old host ids removed from OpenNebula

Details

• Affected Component: Monitor
• Hypervisor: KVM
• Version: 6.4.3

Progress Status

• [ ] Code committed
• [ ] Testing - QA
• [ ] Documentation (Release notes - resolved issues, compatibility, known issues)

[tinova]

transferred to OpenNebula/one https://github.com/OpenNebula/one/issues/6283

[paczerny]

Note for later: I wasn't able to reproduce. It looks like the Host state is not updated in onemonitord on slave. If the Host is deleted and leader changes, the non-updated onemonitord keeps starting monitor agent for non existing host