Proxmox 4.1 HA cluster, no automatic migration VM

Sarge

Renowned Member
Dec 17, 2015
28
2
68
49
Hi,
I installed a cluster of 3 nodes, with iSCSI shared storage.
Then add the VM:100 in HA group. All is ok:

root@pve:~# ha-manager status
quorum OK
master pve (active, Thu Dec 17 22:59:03 2015)
lrm pve (active, Thu Dec 17 22:59:04 2015)
lrm pve2 (active, Thu Dec 17 22:59:03 2015)
lrm pve3 (active, Thu Dec 17 22:59:07 2015)
service vm:100 (pve2, started)
root@pve:~#


But when I rebooting(or shutdown "ipmitool power off") node2(pve2) with VM:100, the VM does not automatically migrate on another node!
In HA status:

Service vm:100(pve2, freese)


The manual says that: Proxmox VE Cluster 4.0 or greater comes with watchdog fencing. This works out of the box, no configuration is required.

root@pve:~# dmesg | grep watchdog
[ 0.146201] NMI watchdog: enabled on all CPUs, permanently consumes one hw-PMU counter.
root@pve:~#


But auto migration is not working.
Manually is OK.
How can I fix the problem?
 

Attachments

  • proxmox.jpg
    proxmox.jpg
    59.2 KB · Views: 24
Last edited:
  • Like
Reactions: chrone
Currently, we freeze VMs when you do a normal shutdown (no migration). But we will change that behavior in the next release, so that
VMs are migrated when you shutdown a node.
 
  • Like
Reactions: chrone
Thank you for answer,
When I pulled the power cable of node, VM is fenced, and migrated after 1.5 min.
But how I can reduce migration time, up to 15 seconds?
Is it possible?

And whether I understand that there is a built-in Proxmox, who understands how the shutdown.
Via IPMI or pulled the power cable?
 
Thank you for answer,
When I pulled the power cable of node, VM is fenced, and migrated after 1.5 min.
But how I can reduce migration time, up to 15 seconds?
Is it possible?

No, that is not possible with watchdog fencing.
 
OK.
Can I use the same fensing in older versions?
Or only a watchdog fencing?
 
OK.
Can I use the same fensing in older versions?
Or only a watchdog fencing?

Watchdog is the only options for now. But we are working on active fencing device support - plan is to add that in 4.2 release.
 
Currently, we freeze VMs when you do a normal shutdown (no migration). But we will change that behavior in the next release, so that
VMs are migrated when you shutdown a node.

Hi dietmar,

I still found the VM is freezed on latest 4.2. How to enable the auto migration if this has been released on PVE 4.2?

Kind regards,

chrone
 
  • Like
Reactions: chrone
That is already implemented:

https://git.proxmox.com/?p=pve-ha-manager.git;a=commit;h=499f06e3d0be1eec0c3c57fdee0a40ce1f5d808c
https://git.proxmox.com/?p=pve-ha-m...;hpb=cde777792f650e268bface995c1ac2f6a54f28ae

We still use 'freeze' state if you do a 'restart' of a node. But a node 'shutdown' stops the VMs now, so the CRM will restart them on another node.

Oh, yes it is, my bad.

What I was asking to ask is about auto live migration for VMs with shared storage should we gracefully reboot a node in HA group.

If I recall correctly, in Hyper-V and XenServer cluster, if we restart a mode, all HA VMs with shared storage will be migrated to other available nodes first before the node itself reboots so the VMs will always be highly available without service disruption or very minimal downtime due to live migration.

Will we see same features in future Proxmox? Or at least the option in Web UI to live migrate all HA VMs which runs on network storage from a single click of a button? :)

/chrone
 
  • Like
Reactions: chrone
We already have that.

Oh yeah, didn't see it hidden under more option. :D Will this migrate only VMs with HA enabled on network storage or non-HA on local storage as well?

I assumed HA enabled VMs only works for network storage, right?

Are you sure 'Migrate all VMs and Containers' only migrates running instances? As I recall it the feature will migrate all VMs and Containers regardless whether they are running or not.

This is what I'm afraid of. We only want to migrate HA enabled VMs which are on shared storage to be auto live migrated to other available nodes in the HA group.

Let me see if I can reproduce this on VirtualBox, afraid to break things up in production. Hehe.

/chrone
 
Last edited:
We already have 'Migrate all VMs and Containers', and I guess we could add additional options there - some filter to choose specific VMs...
 
  • Like
Reactions: chrone
Oh, yes it is, my bad.

What I was asking to ask is about auto live migration for VMs with shared storage should we gracefully reboot a node in HA group.

If I recall correctly, in Hyper-V and XenServer cluster, if we restart a mode, all HA VMs with shared storage will be migrated to other available nodes first before the node itself reboots so the VMs will always be highly available without service disruption or very minimal downtime due to live migration.

Will we see same features in future Proxmox? Or at least the option in Web UI to live migrate all HA VMs which runs on network storage from a single click of a button? :)

/chrone

I too didn't notice the hidden migrate all option under more for a node. I was expecting to see it when right clicking a node in the navigation pane or node list (similar to the options shown when right clicking a vm). Perhaps that could be added? Regardless, I agree VMs should auto live migrate when doing a graceful reboot or shutdown of a node in the HA group or at least pause, notify the user VMs are running and give them the option to abort shutdown/restart of the node until VMs are migrated away.
 

About

The Proxmox community has been around for many years and offers help and support for Proxmox VE, Proxmox Backup Server, and Proxmox Mail Gateway.
We think our community is one of the best thanks to people like you!

Get your subscription!

The Proxmox team works very hard to make sure you are running the best software and getting stable updates and security enhancements, as well as quick enterprise support. Tens of thousands of happy customers have a Proxmox subscription. Get yours easily in our online shop.

Buy now!