HA cause the reboot of node, if you reboot a master.

[is-max]

Hello,

I'v two servers in cluster, one HP and one DELL, boths have ProxMox 4.1-1.

Yesterday I'v configured for test the HA from the webgui, creating groups and adding the virtual machine inside.

Tonight the HP server (wich was in HA tab the one called as "master") had a weird crash on a machine, wich caused the entire system to be unstable,i'v attached the kernel log.


Due to instability, I had to force the reboot trought acpi, it has caused that the DELL server, have rebooted itsel too w\o an apparent error.

When the HP and Dell server has comed back, they have started up all vm's and the system was reliable again.

For ispect the HP machine, and try to understand the issue, I'v moved the VM's from HP to DELL, removed all HA groups, but.. again, when I'v done a simple reboot on the HP server, the DELL server got rebooted again.

I'v checked the log on the Dell server and got that:

Feb 24 10:35:58 vmsrv02 corosync[1299]: [TOTEM ] A new membership (172.16.254.2:24) was formed. Members left: 2
Feb 24 10:35:58 vmsrv02 corosync[1299]: [QUORUM] This node is within the non-primary component and will NOT provide
any services.
Feb 24 10:35:58 vmsrv02 corosync[1299]: [QUORUM] Members[1]: 1
Feb 24 10:35:58 vmsrv02 corosync[1299]: [MAIN ] Completed service synchronization, ready to provide service.
Feb 24 10:35:58 vmsrv02 pmxcfs[1173]: [status] notice: node lost quorum
Feb 24 10:36:00 vmsrv02 pve-ha-crm[1326]: status change slave => wait_for_quorum
Feb 24 10:36:11 vmsrv02 pve-ha-lrm[1334]: status change active => lost_agent_lock
Feb 24 10:36:32 vmsrv02 pvedaemon[1321]: <root@pam> successful auth for user 'root@pam'
Feb 24 10:36:35 vmsrv02 pveproxy[5537]: proxy detected vanished client connection
Feb 24 10:36:57 vmsrv02 watchdog-mux[1047]: client watchdog expired - disable watchdog updates
Feb 24 10:39:05 vmsrv02 rsyslogd: [origin software="rsyslogd" swVersion="8.4.2" x-pid="1076" x-info="http://www.rsysl
og.com"] start

Why that happend? and, btw I'vnt found a way to disable the HA and keep only the cluster active waiting to prepare the third machine.

Thank you so much for your support
regards

 

[dietmar]

Please can you test again with newer 4.2.8 kernel

 

[is-max]

Hi,

Where I can get that kernel? doing apt-get updage&upgrade I didnt see it.

Thank you so much
Regards

 

[dietmar]

Where I can get that kernel?

from standard repositories

 

[is-max]

Hi,

Currently I'vnt bought a subscription, due I'm setting-up the production evirorment.
And I'm a little bit worryed about enable the "non-subscribed" repositories, I could update just the HP machine, or I need to update boths machine in the cluster?

Another info, I could add, I'v tryed to reboot the HP server again (vmsrv01) when it was slave and not master, vmsrv02 hasnt got the wrong reboot.

The watchdog could really have some issue, if the master isnt up, and quorum could not be active?

Thank you
Regards

 

[dietmar]

Please read the docs carefully:

https://pve.proxmox.com/wiki/Proxmox_VE_4.x_Cluster

You need at least 3 nodes for HA.

 

[is-max]

Hi,

yes, I'v understand that, and after that issue, I wanted to disable it, but after deleting all vm's and groups in HA, it still show the HA.

How I could disable? I didnt see it in the wiki.

And anyway, sorry I couldnt understand, why the "slave" server reboot itself, only because it's not quorate.

__EDIT___
OK, I'v understand, why... the "Self-Fencing" act due the system did not have another way to fix it.

Still, I ask help for disable the HA until I could go to take my third server at the DC.
I think, It have something to do with /etc/pve/ha.

Thank you so much for your support.!

Regards

 

[dietmar]

I wanted to disable it, but after deleting all vm's and groups in HA, it still show the HA.

How I could disable?

Remove /etc/pve/ha/resources.cfg should be enough

 

[is-max]

Hi

Thank you so much.
I'v renamed manager_status and resources.cfg and it has disabled the HA
Now on the HA page, show only quorum: OK and not info related to the HA

Thank you so much for your time and help! )

About the kernel issue with the HP, I'll wait to bought the subscription, then let you know if that fix or not.
I think in two weeks.

Thank you again!
Regards