Proxmox V5 - IGMP Snooping querier - Switch replacement

[TwiX]

Hello,

One of our swichs is currently the IGMP snooping querier.

We plan to remove this switch which is an old Cisco 2960. Few days ago I decided to turn this switch off. All my V5 clusters lost quorum and many nodes rebooted.

Is it possible to configure a second switch to be the IGMP snooping querier ? If not, is it possible to settle my nodes to not reboot if they loose quorum (during replacement of that switch and configuring the new one to be the querier) ?

Thanks a lot for your help !

Best regards,

 

[wolfgang]

Hi,

if you like to avoid that the nodes reboot please deactivate the HA.

 

[TwiX]

Thanks understood. You mean remove all Vms or is there a button to switch all on/off ?

 

[wolfgang]

You have to remove them all form HA.
There is no turn off button for this.

But also if all services are removed from HA the lrm is still activated.
The lrm service is responsible to reset the watchdog.
So you have also to reboot the node after the HA Services are removed if the lrm is active.

 

[TwiX]

Thanks. is it possible to shutdown lrm without rebooting the node?

If not it means lot of time due to migrating all vms to another node and rebooting the node.And when node reboot is complete, put some VMs on HA will automatically starts lrm ?

And, without doing these all things, would it be better to declare another igmp snooping querier on my network ?

 

[spirit]

Hello,

Is it possible to configure a second switch to be the IGMP snooping querier ?

Best regards,

yes. you should have 2 igmp queriers on your network at minimum !
Simply enable it on a second switch.
they are an automatic master election based on ip address of management interface.

 

[TwiX]

Thanks, what is the delay between shutdown and election of a new one ? I red somewhere 255s that may explain the reboot of few nodes

 

[spirit]

Thanks, what is the delay between shutdown and election of a new one ? I red somewhere 255s that may explain the reboot of few nodes

on cisco, it's the "querier timeout" command.
I think that the minimum is 60s

 

[TwiX]

And what is the delay for proxmox to reboot when it looses quorum ?

 

[TwiX]

It's me again. Is it safe for my quorum clusters to declare another igmp snooping querier ? I guess new election will begin, and clusters may loose quorum during election ?

Thanks,