[Gluster-devel] [Gluster-infra] r.g.o down?

[Michael Scherer]

Le jeudi 21 avril 2016 à 14:21 +0200, Michael Scherer a écrit :
> Le jeudi 21 avril 2016 à 15:01 +0530, Kaushal M a écrit :
> > On Thu, Apr 21, 2016 at 2:58 PM, Michael Scherer <misc at redhat.com> wrote:
> > > Le jeudi 21 avril 2016 à 14:29 +0530, Kaushal M a écrit :
> > >> On Thu, Apr 21, 2016 at 1:14 PM, Michael Scherer <misc at redhat.com> wrote:
> > >> > Le jeudi 21 avril 2016 à 12:47 +0530, Kaushal M a écrit :
> > >> >> I restarted Gerrit that should have fixed it for now.
> > >> >>
> > >> >> As this has happened several times since the migration, I thought
> > >> >> something wasn't done correctly.
> > >> >> Turns out the gerrit VM is a really-(really)-low end VM with just 1
> > >> >> CPU core and 2GBs of RAM!!
> > >> >>
> > >> >> Michael, was this intentional or did you some how overlook this?
> > >> >
> > >> > I likely overlooked. I reused the same type of VM than the one for
> > >> >
> > >> >>  IIRC,
> > >> >> before the migration the VM was running with a dual-core CPU at least
> > >> >> and more RAM.
> > >> >
> > >> > 4G of ram and 2 cpus. So I propose to go to 6G and still 2 cpu for now ?
> > >>
> > >> Fine with me. We always have room to grow if required.
> > >>
> > >> >
> > >> >> The hypervisor has more than enough resources to run a beefier VM. I
> > >> >> was thinking of bumping the VM to a 4 Cores and 8 GB of RAM (or more).
> > >> >> I'll do this if there are no objections.
> > >> >
> > >> > This requires a reboot however, is this ok ?
> > >>
> > >> Doing this shouldn't take longer that ~5minutes. We could do it any time.
> > >> It would have been so much easier if libvirt allowed hot-plugging cpu
> > >> and ram like disks.
> > >
> > > Seems to be supported in openstack, and in version we use on RHEL 7.
> > >
> > > And virsh has a command setvcpus with a --live switch, and the same for
> > > the ram.
> > 
> > TIL! I've been trying to do this all the time by editing the VM xml
> > config. I believe virt-manager does the same as well, because it fails
> > to add cpus live as well.
> 
> so Kaushal tested and this requires a specific setup for that before. 
> 
> > >
> > > Shall I try later today ?
> > > (worst case, it crash and we reboot)
> > 
> > Sure.
> 
> So i am gonna shutdown the VM and increase ressources now. So 5 minutes
> of downtime.

Of course, things didn't work as planned, since the VM was for some
reason still starting on the xen kernel (likely a kernel upgrade).

So I did wiped all trace of xen, increased the ram, the cpu, planned for
future memory/cpu upgrade if required, and things seems to go well.

Until i found that gerrit was not started at boot correctly. So to be
fixed later, but for now, this is working fine it seems.

-- 
Michael Scherer
Sysadmin, Community Infrastructure and Platform, OSAS


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://www.gluster.org/pipermail/gluster-devel/attachments/20160421/e61d3fb1/attachment.sig>