RHEV - node in maintenance mode automatically joins RHEV after reboot

Latest response

I believe this behavior started with 3.1, and I may be experiencing an irregularity.

When I am planning to reboot my RHEL Hypervisor, I will put it in Maintenance mode (and wait for RHEV to evacuate the VMs).  To my surprise, when the node comes back up - it has automatically rejoined the cluster and will be online.

I would prefer that I have to take it back out of maintenance mode.

Has anyone else noticed this?  Do others share my preference (having persistent maintenance mode through reboots)?

Thanks!

Responses

Is the below assumption correct?

- You are using an RHEL based hypervisor, not RHEV-H.

- Node moved to maintenance mode,  rebooted it. Once it was rebooted, it automatically wet to "Up" status instead of staying in "Maintenance" mode for someone to change the status to Up. Correct?

If the above assumption is correct, then it's not expected. Node should remain in maintenance mode after reboot. Once you confirm,  i will try to reproduce this.

Are you using the latest version of RHEV-H or vdsm packages if using RHEL based hypervisor.

 

I have the same issue and as I thought it is an expected behaviour I've opened an RFE:

https://access.redhat.com/support/cases/00791336

 

You are right, I am using an RHEL host with the latest VDSM packages (vdsm-4.9.6-45.2.el6_3.x86_64). Your procedure is also correct:

1. put the host to maintenance mode

2. restart the host using the "power management" feature from the admin portal

3. the host automatically goes to "UP" state instead of staying in maintenance mode

I am also using RHEL as the Hypervisor, and yes, it transitions to an Up state with no intervention.  The only way to keep it down is to select maintenance mode again while the node is down/rebooting.

I'm not sure what version of VDSM is installed.  We started with RHEL 6.3 and there have been updates applied.  I will update this with the version once I am able.

thx

This is the default behavior. Changing it globally, might not be appreciated by the users who actually rely on it, so I suggest you open a case and request a feature, where hosts' behaviour will be customizeable, per host.