"apt update and apt grade" on HVM (VME) host

[dya]

As a maintenance task, I am planning to periodically update packages using "apt update and apt grade".

What should I do if the VM is running? I would like to know the series of steps to update packages, including what to do.

*This is a machine translation, so I apologize if it is difficult to understand.

[georgek_1]

Hello dya,

You may consider downtime or move the VMs / including VMEssentials Manager to other host before performing apr upgrade.
Running upgrade itself wont trigger automatic reboot but it may be required if kernel updated.

[Learnmoretech]

Hey! No worries—your message is totally understandable.

If your VM is running and you're planning to do apt update and apt upgrade, it's usually safe to run those while the VM is active. But here are a few friendly tips to make sure everything goes smoothly:

1. Make a snapshot or backup of your VM before updating—just in case anything breaks during the upgrade. It’s a good habit.

2. Run:

   sql

   sudo apt update sudo apt upgrade

   You can also use sudo apt full-upgrade if you want to allow it to remove or replace some packages, but that depends on your setup.

3. If the upgrade includes a kernel update or system-level changes, you might need to reboot the VM afterward for everything to take effect.

4. Watch out for any prompts during the upgrade—especially about configuration files. Read carefully before choosing.

So yes, it's okay to run updates while the VM is up, but always have a backup and check if a reboot is needed. Hope that helps!

Let us know how it goes or if you hit any issues

[mikejdunphy]

if you use vme manager gui to place the host in maintenance mode it moves all the vms off for you
and once in maintenance you can patch and reboot to your hearts content and when all done take
out of maintenance

a 'apt upgrade' typically wont install updates like kernel changes that require a update
so instead I always do a apt -y dist-upgrade which gets it all and then reboot to finish them all.

Also the /etc/motd will state if the host needs a reboot or not.

[PeterTzvetanov]

Because of some differences in the way how is to be proceeded and with further clarification of the process to go, I rewrote now the reply.

Peter Tzvetanov

Support Engineer

Peter.Tzvetanov@hpe.com

Technical Support Phone Numbers

HPE Operations

You can manage your cases on HPE Support Center

[mikejdunphy]

thats pretty ugly and defeats the whole purpose of having a hypervisor and a HA cluster.
I have already done the process a few times and never lost the whole thing.

Even if it did .. its the same result of all the vms going down and not being available just
via a crash instead of a planned maintenance..

[davydotcom]

If performing periodic security updates with apt-get upgrade, Please first ensure the host is in maintenance mode . This means no vms are running on the host. Perform the upgrade operations. Once these are completed, run `sudo pcs status` and confirm pacemaker is healthy after the upgrade. If it is not, reboot the host and check again. If it still is not, run a `sudo pcs resource cleanup`.  After you have confirmed everything is healthy, you can leave maintenance mode from the host details page in the vme manager.

[dya]

Is maintenance mode an unsupported operation for a GFS2 cluster?

If so, shouldn't it be impossible to enter maintenance mode?

[mikejdunphy]

Works for me, I have no issues with it.

[signature_406722384]
Mike Dunphy, Systems Engineer
Information Technology Group
Oregon Health & Science University
Mobile: 541.230.8629 | Email: dunphy@ohsu.edu

[Arnout_Verbeken · Accepted solution]

"Maintenance mode " is supported in a GFS2 cluster.

Maintenance mode will move all running VM's to other nodes in the cluster, but the host itself will still be a node of the Pacemaker cluster.
In my lab, I constantly have 1 node in Maintenance (because I do not really trust that hardware), but that node is an "online" node in the pacemaker cluster. So the node is part of the VME Cluster but in maintenance, part of the pacemaker cluster and fully participating there, but does not run any VM's.

There is no need to shut down all VM's when putting a host in Maintenance.

You set a host in Maintenance via the VME GUI.  Once in maintenance, you can patch, reboot,...
You can put multiple servers at the same time in Maintenance mode, but you need to make sure that you always have a majority online in the pacemaker cluster.
That's why I recommend to reboot only 1 host at a time.
Always very with "pcs status" that your pacemaker cluster in fully online before rebooting a host.

[PeterTzvetanov]

As before declared, rewrote the post because of some further process investigations internally.

[dya]

- Could you please explain the first screenshot? Is it from a host in maintenance mode? Does this only happen when in maintenance mode?

- You say "And the others died with the GFS2 too:" - is this due to maintenance mode? Is this an issue that occurs in GFS2 clusters, regardless of HVM itself or HVM maintenance mode? I'm curious to know the underlying problem that's causing this issue.

[PeterTzvetanov]

Revising the post for further 

[dya]

It seemed that there was a risk inherent in GFS2 itself.

[cbunge]

I would disregard what Peter has stated and we will correct internally so everyone is on the same page.

Others have correctly highlighted that maintenance mode (mm) will move vms to other hosts and you are safe to do apt actions on the host.

While yes, GFS2 relies on Pacemaker and Corosync and therefore a quorumed cluster for storage, a single host [for example] having an issue with an apt upgrade will not terminate your GFS2 cluster as long as you maintain minimum node counts (which is (number of hosts / 2) + 1). So in a 3 node cluster you must maintain 2 healthy hosts, i.e - maintenance 1 at a time.

[dya]

I understand that this is not a bug found in HPE Morpheus VM Essentials, but rather the same concept as when forming a cluster with pacemaker.