[OpenAFS] Fw: drbd or nbd with openafs?

[Benko]

This is a multi-part message in MIME format.

--Multipart=_Thu__22_Jun_2006_14_01_55_+0200_3AtkvJEr3y=bkRXU
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

This is somehow a reply to a message from 2003(link is in my reply after
the quote). Brent, sorry for sending this to you twice today, but
i thought it would be good to share it with the list...

> I don't have anything written up for it, and it hasn't really been
> tested with active users, but I do believe I have a working OpenAFS
> server setup with drbd (enbd would work fine, too).
> 
> There were a few tricks involved in setting it up.
> 
> 1) I made sure that only my failover IP was being registered with the
> volume database using the NetInfo file (just put in the failover IP
> address in /usr/afs/local/NetInfo).
> 
> 2) The big trick was forcing my takeover node to use the the failover
> IP address (which is an IP alias) as the source address on outgoing
> packets from that interface.  See the attached heartbeat script (goes
> in /etc/ha.d/resource.d).  The method I used is probably
> Linux-specific...
> 
> 3) Finally, I had to make sure the machines were using the same ID (I
> used a hexeditor on the /usr/afs/local/sysid file, but maybe someone
> else has a better method).
> 
> Both machines are database servers, with only the active file server 
> running "fs" at any given time.  They both have "fs" defined but not
> set to run automatically and bos is set to never restart the services 
> (otherwise it would do the wrong thing).
> 
> I think that was it, but I may have missed something.  I might be
> able to help if you get stuck.
> 
> Thanks,
> 
> Brent Nelson
> Director of Computing
> Dept. of Physics
> University of Florida

Hello Brent!

I'm currently trying to set up a testcluster with heartbeat, drbd and a
openafs-fileserver. Heartbeat and drbd are running beautifully, but
setting up the openafs failover seems to be quite difficult. 

I've found your post from 23 June 2003 in the openafs mailinglist
(http://lists.openafs.org/pipermail/openafs-info/2003-June/009760.html),
which was the only information to a working setup(and how to do it) i've
found so far.
Unfortunately i've not been able to copycat your method yet. How did
you change the sysid-file? This file is very small in my case(only one
line) and i can't see something like a ID when i open it with a
hexeditor, there are only many empty parts and a few characters(i've
attached my sysid-file).

You also wrote:
> Both machines are database servers, with only the active file server 
> running "fs" at any given time.  They both have "fs" defined but not
> set to run automatically and bos is set to never restart the services 
> (otherwise it would do the wrong thing).

so you only start/stop the fileserver on failover, the other servers run
all the time on both nodes? Sorry if this question is trivial, i'm still
missing understanding of openafs unfortunately (though i've already
read a lot, guess you know what i mean ;-))

Finally - do you still use the same method for the openafs-failover? Or
have you found better ways to do it?

It would be great to hear from you!

Thanks in advance from austria
Christian Benke



--Multipart=_Thu__22_Jun_2006_14_01_55_+0200_3AtkvJEr3y=bkRXU
Content-Type: application/octet-stream;
 name="sysid"
Content-Disposition: attachment;
 filename="sysid"
Content-Transfer-Encoding: base64

zLuqiAEAAAAAl0DSdjYUmrF+AQAAf6p3AwAAAAoBY5oKAWOdwKgBAg==

--Multipart=_Thu__22_Jun_2006_14_01_55_+0200_3AtkvJEr3y=bkRXU--