File-backed VBDs Migration help

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

File-backed VBDs Migration help

Jonathan Wheeler
Hi All,

My Apologies if this has been asked/answered already, premptively in my
defense, I have already spent 3 hours in the mailing list archives and
with google.

Why is it, that I can't use a local file based VBD, and then migrate it
to another host?
I can use a file based VBD and migrate, if that file is on a nfs mount,
that both servers use, but I really don't want to go this way. Here's why:

This is for a small company setup, so there is no SAN/ISCSI - and no
chance of getting it either.
I'm wanting to setup a 3 way frontend cluster, which then backends onto
a NFS server for mail/web/blah. All outside the scope of Xen...

The advantage of course with Xen is that I could move the DomU's between
physical servers for maintenence and scheduled downtime without any real
downtime - hurrah.

I have a 100Mbit network, and if I'm forced to run all my guests off the
one NFS server, that's only 33Mbit each and that's *before* they domUs
actually start trying to do anything (ie serving up web pages/ftp/mail
etc). Real world math of course will work out differently, but there's
no arguing that having the three domUs running across the network is
going to be slower then running them locally.

Further - if the NFS server goes away for whatever reason, so do *ALL*
my front end servers - pretty much defeating the point of the cluster.

So running the guests off this one NFS server really doesn't seem like a
good idea. Running them off a local file/partition dramatically
increases I/O and leaves the network free for doing it's job. However I
can't do a migration (live or otherwise) using this method, from my
testing so far.

Am I missing something, or does anyone have some suggetions for my
situation?

Many thanks,
Jonathan.

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: File-backed VBDs Migration help

James Bulpin-2
Jonathan,

If you only want migration for maintanence purposes then you could
probably put up with a small downtime during the migration. How about
having your "migrate" sequence as:

xm save mydomain /path/mydomain.save
rcp [or whatever] /path/mydomain.save otherhost:/path/mydomain.save
rcp /path/mydomain-vbd.img otherhost:/path/mydomain-vbd.img

then on otherhost:

xm restore /path/mydomain.save

(Disclaimer: I've not tried this but I can't see any reason why it
wouldn't work.)

Downtime would be dictated by the size of your save and vbd files and
the speed of the network between the hosts.

James

Jonathan Wheeler wrote:

> Hi All,
>
> My Apologies if this has been asked/answered already, premptively in my
> defense, I have already spent 3 hours in the mailing list archives and
> with google.
>
> Why is it, that I can't use a local file based VBD, and then migrate it
> to another host?
> I can use a file based VBD and migrate, if that file is on a nfs mount,
> that both servers use, but I really don't want to go this way. Here's why:
>
> This is for a small company setup, so there is no SAN/ISCSI - and no
> chance of getting it either.
> I'm wanting to setup a 3 way frontend cluster, which then backends onto
> a NFS server for mail/web/blah. All outside the scope of Xen...
>
> The advantage of course with Xen is that I could move the DomU's between
> physical servers for maintenence and scheduled downtime without any real
> downtime - hurrah.
>
> I have a 100Mbit network, and if I'm forced to run all my guests off the
> one NFS server, that's only 33Mbit each and that's *before* they domUs
> actually start trying to do anything (ie serving up web pages/ftp/mail
> etc). Real world math of course will work out differently, but there's
> no arguing that having the three domUs running across the network is
> going to be slower then running them locally.
>
> Further - if the NFS server goes away for whatever reason, so do *ALL*
> my front end servers - pretty much defeating the point of the cluster.
>
> So running the guests off this one NFS server really doesn't seem like a
> good idea. Running them off a local file/partition dramatically
> increases I/O and leaves the network free for doing it's job. However I
> can't do a migration (live or otherwise) using this method, from my
> testing so far.
>
> Am I missing something, or does anyone have some suggetions for my
> situation?
>
> Many thanks,
> Jonathan.
>
> _______________________________________________
> Xen-users mailing list
> [hidden email]
> http://lists.xensource.com/xen-users

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xensource.com/xen-users