Re: [Xen-users] debian squeeze, xen and a very big problem.

[Simon Walter <simon.walter@xxxxxxxxxxxxxxxxxx>]

On Friday, October 14, 2011 11:59:23 PM Florian Heigl wrote:
> "You need to run your own test servers if you want a working update
> since we don't test stuff like live migration (that is most basic
> functionality)" makes me chuckle a bit. 2 test servers can buy
> helluvah support contract, where the issues would probably be
> addressed quickly. Shouldn't it be that the user tests anything that
> is related to oddness of his own setup and a distro tests all that is
> basic functionality?
> 
> If a upgrade from minor release A to minor release B fails like that
> it's perfectly OK to come forward and say "OK that must be a
> regression, sorry. Please open a bug immediately" instead of yet
> another round of
> Q: Xen issue in Debian
> A: "blame and flame the user"

Totally agree. However, we cannot trust very much in this industry. It's just 
the way it is. With my Xen cluster I would do so:

DomU upgrades:
* Clone the machine and perform the upgrade on the clone and test.
* If it goes well, clone the machine for backup, upgrade the production 
machine.
* If things go wrong, you have a back up.

There may be a problem with loss of data with the above method. If everything 
goes well for a few days on the new system. Data is modified. The backup is 
useless unless you also consider backups of just the data partition/volume and 
try use that with the old system.

Dom0 upgrades:
* Migrate DomUs off of Dom0 being upgraded.
* Copy them back possibly starting with a test machine.
* Test them.
* Migrate them back.

Xen itself empowers us to do safer upgrades to our machines. It is one of the 
reasons for using it.

If you can only afford one Dom0, then things get trickier, but it's still 
possible to be safer than doing upgrades to a production server w/o testing.

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users