Jump to content


Photo

Upgrading 6.5 with Network Bonds and VLANs to xenserver 7

Started by G. Reinicke , 10 January 2017 - 08:04 PM
22 replies to this topic

Best Answer Jiri Cerny , 11 January 2017 - 10:20 AM

Hello Götz,

if you'll use Rolling Pool Upgrade (and I suppose yes, because want VMs keep running), it will test your system a and if you haven't any VM on local storage, RPU mark system ok for change partition layout. So RPU make "touch /var/preserve/safe2upgrade".

Manually creating "safe2upgrade" is just workaround for people, who upgrade manually without RPU procedure. 

If anything goes wrong, XS7 installer restore process should restore original partition layout.

 

Jiri 

G. Reinicke Members

G. Reinicke
  • 72 posts

Posted 10 January 2017 - 08:04 PM

Hi,

 

we havs a pool of three xenserver V 6.5sp1 with a couple of VLANs on 10G Networkbonds.

 

There are no local SRs used every VM is on a NFS SR.

 

The Partitions are GPT. 

 

So far an upgrade path looks good, but I could not find any suggestions or notes on the bonds and VLAN configuration.

 

Can someone comment on this?

 

Thanks and regards . Götz

 



Alan Lantz Members

Alan Lantz
  • 6,809 posts

Posted 10 January 2017 - 09:02 PM

When you upgrade all of your settings will remain intact. A fresh upgrade you would need to reset your networks, but an upgrade you should be okay. Make sure your hardware is on the HCL or someone can vouch that your hardware will work with 7.

 

--Alan--



Tobias Kreidl CTP Member

Tobias Kreidl
  • 18,052 posts

Posted 10 January 2017 - 11:47 PM

If you do a rolling pool upgrade, yes, and keep the disk layout. If not, it's almost as much work as a clean install (as I have just experienced with two pools going from 6.5 SP1 to 7.0), and in any case, you always want to document how each server is layed out. I take snapshots of the network config settings in XenCenter as an aid. I also do a backup of the metadata since the external storage needs to be re-attached, plus vacate anything on any local SRs since that's going to get lost with a clean install.

 

-=Tobias



G. Reinicke Members

G. Reinicke
  • 72 posts

Posted 11 January 2017 - 07:58 AM

Hi Alan, hi Tobias,

 

thanks for your feedback. So same procedure as usual. It should work; I'd like to go with an rolling pool upgrade to keep the VMs online we need; I'd like to see if that works in real live too.

 

NIC and CPU are in the HCL for 6.5 and 7.

 

@Tobias how do you do the network snapshot; taking a screenshot :)

 

An other point is; during the upgrade I'd like to upgrade to the new partition layout for the hosts as sometimes logs did fill up the 6.x layout :(

 

Any notes on the "touch /var/preserve/safe2upgrade" way?

 

Regard . Götz



Jiri Cerny Members

Jiri Cerny
  • 116 posts

Posted 11 January 2017 - 10:20 AM

Hello Götz,

if you'll use Rolling Pool Upgrade (and I suppose yes, because want VMs keep running), it will test your system a and if you haven't any VM on local storage, RPU mark system ok for change partition layout. So RPU make "touch /var/preserve/safe2upgrade".

Manually creating "safe2upgrade" is just workaround for people, who upgrade manually without RPU procedure. 

If anything goes wrong, XS7 installer restore process should restore original partition layout.

 

Jiri 


Best Answer

Alan Lantz Members

Alan Lantz
  • 6,809 posts

Posted 11 January 2017 - 03:28 PM

The only thing is with a rolling pool upgrade I had to use media, I couldn't get the ftp/http option to work. You definitely want the new partition scheme layout if at all possible. 

 

 

--Alan--



Tobias Kreidl CTP Member

Tobias Kreidl
  • 18,052 posts

Posted 11 January 2017 - 04:48 PM

My experience was that a rolling pool upgrade did not work at all if you (1) used fixed media, and (2) wanted the new partition layout. The Rolling Pool upgrade wizard neglects to tell you this isn't going to work and ends up having the pool get decimated in the process.



Alan Lantz Members

Alan Lantz
  • 6,809 posts

Posted 11 January 2017 - 04:55 PM

Well, funny you should mention that Tobias. I had to install my XenServer 7's fresh to get the new partition layout. I assumed that the rolling pool upgrade caught something it didn't like so it stuck with the old partition scheme. Maybe thats just the way it works. Anyone with a decent size drive/raid should go to the new layout which I guess means new installs or upgrade with media and bypassing the rolling pool upgrade.

 

--Alan--



G. Reinicke Members

G. Reinicke
  • 72 posts

Posted 11 January 2017 - 07:53 PM

Hi, O.K. now we are back on might or might not work .... 

 

Tobias, what do you mean by "fixed media"?

 

From the official Doc I read, that a rolling pool upgrade will migrate automatically to the new disk layout if you have no VDs on the local SR and enough space left.

 

We have a 200 GB SSD Hardware Raid 1. 

 

/G



Alan Lantz Members
  • #10

Alan Lantz
  • 6,809 posts

Posted 11 January 2017 - 09:52 PM

A rolling pool upgrade should do the new disk layout, but for whatever reason mine didn't. I had to do a new clean install before the disk partition layout changed. Have no idea why and I didn't try to trick it with touching a file either. 

 

--Alan--



Tobias Kreidl CTP Member
  • #11

Tobias Kreidl
  • 18,052 posts

Posted 11 January 2017 - 11:46 PM

Mine didn't either, which was a huge disappointment. The issue is that the new disk layout requires 6 partitions, so not enough to preserve the old one. Why don't they just package up the whole previous installation and allow you to put it somewhere and make it possible to retrieve it as a restore option??? I don't see how this can ever work and if you run the rolling pool upgrade, you're never warned that this may become an issue. In fact, even if you do the full install from media, that's the end of your pool as it loses all the association and the rolling pool upgrade process fails miserably. Not at all happy with that and it cost a lot of extra time to re-create network links, re-attach storage, etc. Thankfully, I had good backups of VMs as well as the metadata.

 

-=Tobias



G. Reinicke Members
  • #12

G. Reinicke
  • 72 posts

Posted 12 January 2017 - 08:37 AM

OK, that did not makle me feel better in any kind, on the other hand there are reports that this should work .... a 50% chance on success and failure.  :huh:

 

So may be I'll shutdown the VMs and try the rolling pool upgrade .... and if that fails install the three hosts and rebuild a pool.

 

If that is the case, how can I reconnect/enable the shutdown VMs on the pool? 

 

Thanks for suggestions on this. B)

 

Regards . Götz



Jiri Cerny Members
  • #13

Jiri Cerny
  • 116 posts

Posted 12 January 2017 - 01:14 PM

Hello Götz,

you can backup metadata via xsconsole -> Backup, Restore and Update -> Backup Virtual Machine Metadata and specify SR for backup. That's all.

 

Make sure, that SR is detached from original pool or original pool is destoryed at all.

 

On new/another pool run xsconsole -> Disks and Storage Repositories -> Attach Existing SR and specify ty of SR and technical information. Then Backup, Restore and Update -> Restore Virtual Machine Metadata and specify SR attached before. You could restore only VMs of that SR or all VMs metadata (whole pool).

 

Edit:

"Make sure, that SR is DEtached..." Sorry for typo



Jiri Cerny Members
  • #14

Jiri Cerny
  • 116 posts

Posted 12 January 2017 - 01:33 PM

Tobias:
I think you have restore option. You can yet downgrade XS from 7 to 6.5(2). XS7 installer/restorer take care of partitioning so it will restore old partitioning scheme. Of ourse, it a little bit denies purpose of RPU because VMs cannot be migrated from 7 to 6. So you stuck on partly upgraded pool and you have to shut down VMs. Than you are on same journey as people who upgrade pool manualy with install CD.

That's reason, why I always do RPU with all VMs powered down.

 

If you had your post in mind different way, excuse me for misunderstanding;)



Alan Lantz Members
  • #15

Alan Lantz
  • 6,809 posts

Posted 12 January 2017 - 02:17 PM

RPU with VM's powered down is an excellent suggestion if one can do that. Some of my pools I can and some I  can't. I'll have to remember that for the upcoming 7.1.

 

--Alan--



Tobias Kreidl CTP Member
  • #16

Tobias Kreidl
  • 18,052 posts

Posted 12 January 2017 - 03:26 PM

I never even got an option whether or not I could retain the old disk layout or not! I suspect it might be because of the stupid Dell Utility partition it put in during some initial install that isn't wiped out on all our servers, and therefore you don't have enough spare partitions available (just a theory). Not happy about the whole process and poor info given on RPU on XenCenter. Will probably just bite the bullet and stick to full installs to avoid any crap partitions being preserved; they are all gone with the full install, thankfully.

The other issue is that if you install from disk and something goes bad, your pool is messed up. Finally, always, always disable HA before you start an upgrade/downgrade!

 

-=Tobias

 

P.S. By "fixed media" I mean doing the install with a CD/DVD-ROM as opposed to NFS or http.



Alan Lantz Members
  • #17

Alan Lantz
  • 6,809 posts

Posted 13 January 2017 - 04:08 PM

I still think the pool master green light should flash to let us know HA is enabled.

 

--Alan--



Tobias Kreidl CTP Member
  • #18

Tobias Kreidl
  • 18,052 posts

Posted 13 January 2017 - 06:04 PM

HA-Lizard handles screw-ups with HA way better than Citrix' own HA product (which was purchased from some 3rd party vendor ages ago and likely not touched since). ^_^  It is crazy that it will let you shut down the pool master with HA enabled without a warning or making you disable HA first or via the CLI, requiring a "--force" option. :angry:



G. Reinicke Members
  • #19

G. Reinicke
  • 72 posts

Posted 08 February 2017 - 02:07 PM

Hi all,

 

just to give some feedback; I started a NFS based rolling pool upgrade and all prechecks where ok.

 

But after the first automatic reboot the first server and booting to the upgrade image, the NFS repository could not be found.

So the install failed and aborted and rolled back to the 6.5 Pool.

 

I did than the manual reboot-upgrade-procedure with a CD and everything went well.

 

The local system disk where repartitioned. Well done.

 

My guess regarding the failing NFS is, that my VLAN Nic Bonding setup is not working in that stage.

 

/G



Alan Lantz Members
  • #20

Alan Lantz
  • 6,809 posts

Posted 08 February 2017 - 02:13 PM

There has been a lot of reports of this. I think that network initialization is happening later in the boot process and networking isn't available when needed for RPU.

 

--Alan--