RE: High Availability Samba file Server- Piranha FOS+DRBD

On Fri, 2003-07-25 at 09:22, Sumesh wrote:

>  Sorry, I am not clear. Can I understand that , at the moment I can't go for Piranha FOS 
>  for a HA Samba Fileserver failover  solution ?
> 
> Kindly clarify
> 
> 
> Thanks&regards
> Sumesh
> <clip>

In its default configuration, piranha uses a test to http to see if the
other box is online. If you needed more robust testing like actually
making sure you could get to the samba shares, you would have to create
the test scripts yourself. However, piranha straight out of the box does
not support failover script testing, it does for lvs which is where you
have a virtual server that dishes out requests to slave boxes. That
isn't what you are looking for though, if I read you right. I think you
are trying to get what I have running, where you have pairs of boxes, a
master and a slave. When the master dies, stops responding, or other
malady, the slave detects it and takes over. To the outside world it
appears as if the box is still up. Red Hat out of the box, does not
really support this scenario, even though they have built in FOS, they
go more for clustering and indeed seem to have abandoned piranha in
favour of cluster manager. 
In my configuration, I have fw-fw, app server-app server, and
postgres-postgres pairs. The firewalls hold the external shared address
and failover that address on a firewall shutdown. The firewalls forward
requests to the app server shared address. If the app is down then it
fails over to the backup server. This is one place where just testing
for http access would not work as this box uses jboss also and that can
hang while http still runs,  so more intricate testing is involved. I
found out at this time that FOS did not support running external
scripts, only in LVS mode. Fortunately, Sebastien had patched piranha to
allow for this. It works great, but the caveat is that it isn't stock
Red Hat. I have forwarded Sebastien's patches on to Red Hat for possible
inclusion in future versions of piranha if they have not abandoned it
all together. Oh, for the postgres servers we run an external script
that actually attempts to retrieve data to ensure that the database is
indeed up and available. 
I don't know why Red Hat didn't include this capability in the stock
piranha. Hopefully, they will. 
Hope that didn't muddy the waters even further.
Ed