KVM HA feature too slow

[baltazorbest]

Hey!
I have the Cloudstack version 4.19.1.3 with 3 management hosts and 5 compute kvm nodes. I need to figure out how to enable HA in my infrastructure.
My primary and secondary storages are NFS.
I enabled HA on KVM hosts and set kvm provider KVMHAProvider.
Then I created the compute offering with HA feature and a few instances with HA offering.
After that I turned off one kvm host (to simulate sudden power cable unplug) and expected that the server would change its status to ‘down’ soon but it happened roughly 30 minutes later.

I changed several global settings and restarted cloudstack-management one by one but it didn’t help.
The changes were:

• alert.warm: 1800 -> 120
• ping.interval: 60 -> 10
• kvm.ha.degraded.max.period: 300 -> 120

What should I do to force the ‘failed’ host to change its status to ‘down’ after 2 or 3 minutes and make the instances from this host to start on other hosts?
Thank you in advance.

[DaanHoogland]

@baltazorbest , there are several other settings , like kvm.ha.activity.check.interval, kvm.ha.activity.check.max.attempts, etc. search for ".ha." in the global settings tab to find more.

Keep in mind that you need to avoid the chance on a split brain scenario. take care

[rp-]

This #9677 might also be interesting for you.

I think with your installed version, you can only lower the host down if you set the wait configuration lower.

[baltazorbest]

Hey.

Thanks.
After changing the wait parameter from 1800 seconds to 20 seconds, I see a difference, and now the host changes its status to Down after 5 minutes.