Hi All,
I have a production deployment of ACS managing three XenServer Clusters
(XenServer pools of 6 hosts each) in two different Zones. I now find myself in
the position of needing to do a major version upgrade of those hosts. Happily I
have a ACS lab managing a XenServer cluster running the same (old) version of
XenServer that I can practice on.
I have plenty of practice operating ACS to “quiesce things” for XenServer
patches (start with the pool master, move guest VMs off, put that host into
maintenance mode, unmanage the cluster, patch the host, then reverse & move
onto the next host with the same steps except we don’t bother
w/un-managing/re-managing the cluster), but as I understand a XenServer version
upgrade backs up the whole original XenServer installation to another partition
and makes a clean installation of XenServer on the original partition (the
problem there being that when the upgraded XenServer boots up all the ACS
provided/copied scripts are not there & ACS can’t manage the host).
So not much of an ask here (OK maybe at the end – have I missed something
obvious or doing anything foolish?), I wanted to share a bit research & lay out
a set of steps that I think will work to get a pool of XenServers in a cluster
upgraded and end up in a place where ACS is happy with them.
Bear with me it’s a little long,
1. In XenCenter – if HA is enabled for the XenServer pool, disable it
2. Stop ACS management/usage services
3. Do MySQL database backups
4. Start ACS management/usage services
5. Start with the pool master.
6. In ACS – Migrate all guest VMs to other hosts in the cluster
7. In ACS – Prepare to put the pool master into maintenance mode (so no new
guest VMs)
* A caveat here related to this item I found when researching –
https://www.shapeblue.com/how-to-upgrade-an-apache-cloudstack-citrix-xenserver-cluster/
i. A
recommendation is made here to edit
/etc/cloudstack/management/environment.properties
ii. And set
manage.xenserver.pool.master=false
iii. And
restart CloudStack management services
iv. BECAUSE if
one didn’t I understand CloudStack WOULD force an election for another host to
become the pool master (which is “bad” as ASCs is configured to speak to the
currently configured pool master)
* HOWEVER THIS MAY NOT BE NECESSARY
i. Found a
thread titled “A Story of a failed XenServer Upgrade” here –
http://mail-archives.apache.org/mod_mbox/cloudstack-users/201601.mbox/browser
ii. At the
end of the thread Paul Angus states that Geoff’s ShapeBlue blog article was
written in the ACS 4.3 era and that ACS’ behavior “used to be that putting a
host that was the pool master into maintenance would cause CloudStack to force
an election for another host to become pool master - stopping you from then
upgrading the active pool master first. There wasn't an 'unmanage' button at
the time.”
iii. The
implication, (in my humble estimation) is that, today, one no longer needs to
make these edits to /etc/cloudstack/management/environment.properties
1. In ACS – Put the pool master into maintenance mode (so no new guest VMs)
2. In ACS – Un-manage the cluster
3. NOW – Upgrade the XenServer pool master to the latest release
* Exercise left to the reader…(I’ll share what I end up doing)
4. In XenServer – Do the following (found this nugget in a thread last
November about XenServer upgrades & how to re-setup the host – thanks Richard
Lawley!)
* Remove this host tag: xe host-param-remove uuid=HOSTUUID
param-name=tags
param-key=vmops-version-com.cloud.hypervisor.xenserver.resource.XenServer650R
* This makes management server set the host back up, presumably since
ACS has credentials to the host in the database it can copy all the scripts back
5. In ACS – Re-manage the cluster
6. In ACS – Exit maintenance-mode for the newly upgraded host
7. In ACS – Observe that the newly upgraded host is “Enabled” and “Up” in
the UI (Infrastructure --> Hosts)
8. In ACS – Testing (e.g. move an existing router/VM to the upgraded host,
create new networks/VMs on the upgraded host)
9. Rinse & repeat with the remaining XenServer pool members in the ACS
cluster.
* WITH THIS EXCEPTION: No need to manipulate management of the cluster
in ACS
10. In XenCenter – if HA is required re-enable it
Now all of this completely bypasses steps that are spelled out here (which I
*suspect* might now be “old”?):
*
http://docs.cloudstack.apache.org/en/latest/installguide/hypervisor/xenserver.html#upgrading-xenserver-versions
which asks one to run a lot of scripts (cleaning VLANs, preparing for upgrade
etc.) in addition to copying files from the management server to the XenServer
host (to set it back up again for ACS). I’m really hoping that step 11 saves me
from that.
So (asks after all):
1. Is this a reasonable/viable plan?
2. Is my take on step 7 correct?
Thanks for taking a look,
David
David Merrill
Senior Systems Engineer,
Managed and Private/Hybrid Cloud Services
OTELCO
92 Oak Street, Portland ME 04101
office 207.772.5678<callto:207.772.5678>
http://www.otelco.com/cloud-and-managed-services
Confidentiality Message
The information contained in this e-mail transmission may be confidential and
legally privileged. If you are not the intended recipient, you are notified
that any dissemination, distribution, copying or other use of this information,
including attachments, is prohibited. If you received this message in error,
please call me at 207.772.5678<callto:207.772.5678> so this error can be
corrected.
