You are viewing a plain text version of this content. The canonical link for it is here.
Posted to builds@apache.org by Lance Albertson <la...@osuosl.org> on 2019/04/02 23:33:37 UTC

[osuosl-openpower] Planned outage for POWER9 VMs - Wednesday April 3, 2019 9:00AM PDT (1600 UTC)

All,

I need to have a short outage for all POWER9 virtual machines running on
our cluster. I've been doing kernel upgrades on the hypervisors and I ran
into an interesting problem on the POWER9 machines.

For whatever reason, I'm not able to complete a live migration between a
POWER9 hypervisor running the 4.14 kernel onto a machine which has the
newer 4.19 kernel. I've confirmed that once the a VM has been moved to a
machine running the newer kernel, live migrations work between nodes
running the newer kernel. I did not run into this problem with the POWER8
nodes.

My plan is to issue a "migrate" command for each POWER9 VM running at 9AM
PDT (1600 UTC) which will gracefully shutdown the VM and  power it up on
another server. I expect the outage should last no more than a few minutes
per VM.

If this is a problem for your project, please let me know ASAP.

Thanks-

-- 
Lance Albertson
Director
Oregon State University | Open Source Lab

Re: [osuosl-openpower] Planned outage for POWER9 VMs - Wednesday April 3, 2019 9:00AM PDT (1600 UTC)

Posted by Lance Albertson <la...@osuosl.org>.
All,

This I did this earlier this morning and neglected to notice that there
were a set of about 10 VMs that did not start properly upon migrating them.
I just went through all of them and they should be up now. I think the
problem was related to an issue with the neutron-linuxbridge-agent not
working properly initially and thus resulted in a failed VM startup on the
other node.

Everything should be back to normal now.

Thanks-

On Tue, Apr 2, 2019 at 4:33 PM Lance Albertson <la...@osuosl.org> wrote:

> All,
>
> I need to have a short outage for all POWER9 virtual machines running on
> our cluster. I've been doing kernel upgrades on the hypervisors and I ran
> into an interesting problem on the POWER9 machines.
>
> For whatever reason, I'm not able to complete a live migration between a
> POWER9 hypervisor running the 4.14 kernel onto a machine which has the
> newer 4.19 kernel. I've confirmed that once the a VM has been moved to a
> machine running the newer kernel, live migrations work between nodes
> running the newer kernel. I did not run into this problem with the POWER8
> nodes.
>
> My plan is to issue a "migrate" command for each POWER9 VM running at 9AM
> PDT (1600 UTC) which will gracefully shutdown the VM and  power it up on
> another server. I expect the outage should last no more than a few minutes
> per VM.
>
> If this is a problem for your project, please let me know ASAP.
>
> Thanks-
>
> --
> Lance Albertson
> Director
> Oregon State University | Open Source Lab
>


-- 
Lance Albertson
Director
Oregon State University | Open Source Lab

Re: [osuosl-openpower] Planned outage for POWER9 VMs - Wednesday April 3, 2019 9:00AM PDT (1600 UTC)

Posted by Mo Zhou <lu...@debian.org>.
Hi,

Thank you all for offering such nice VM service! I've just read the LWN news.
I'm fine with even longer offline time since Debian related development is quite flexible.

On Tue, Apr 02, 2019 at 04:33:37PM -0700, Lance Albertson wrote:
> All,
> 
> I need to have a short outage for all POWER9 virtual machines running on our
> cluster. I've been doing kernel upgrades on the hypervisors and I ran into an
> interesting problem on the POWER9 machines.
> 
> For whatever reason, I'm not able to complete a live migration between a POWER9
> hypervisor running the 4.14 kernel onto a machine which has the newer 4.19
> kernel. I've confirmed that once the a VM has been moved to a machine running
> the newer kernel, live migrations work between nodes running the newer kernel.
> I did not run into this problem with the POWER8 nodes.
> 
> My plan is to issue a "migrate" command for each POWER9 VM running at 9AM PDT
> (1600 UTC) which will gracefully shutdown the VM and  power it up on
> another server. I expect the outage should last no more than a few minutes per
> VM.
> 
> If this is a problem for your project, please let me know ASAP.
> 
> Thanks-
> 
> --
> Lance Albertson
> Director
> Oregon State University | Open Source Lab 

> _______________________________________________
> openpower mailing list
> openpower@osuosl.org
> https://lists.osuosl.org/mailman/listinfo/openpower

_______________________________________________
openpower mailing list
openpower@osuosl.org
https://lists.osuosl.org/mailman/listinfo/openpower