[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Piranha & samba

I did not use Piranha, but it should certainly be possible.

Your biggest issue is synchronized storage.  My company set up a test-bed
for the Atlanta Linux Showcase that used a dual-initiator SCSI attached (dual
attached, obviously) RAID subsystem for both failover samba and NFS.

I did not do the samba part, but it appeared to be fairly easy.  I think
since SMB is stateful, the connection will appear to hang after failover,
and you will have to re-establish the connection (ie break the connection
and connect again, from the client - I am not proficient in Windows so
I can't give you the detailed steps).

NFS was even easier, the failover was transparent, but you have to get the
device names for the disk/RAID identical on both master and failover to
avoid getting stale file handles.  NFS uses the device and location of the
data on the disk to construct the filehandles, so it will only work properly
with identical disks on identical devices (logically, ie same device name,
same major and minor numbers, etc) on both systems.  As I said, we
accomplished this by using a dual attached SCSI RAID subsystem, and
by paying attention to device creation.

We did not use Piranha for this, we just used heartbeat, which worked
quite well.  I don't see any reason that Piranha couldn't do the job
as well, if the disk storage is set up properly.



One thing you will have to consider for a production environment is
how to make absolutely certain that both systems cannot end up mounting
the disks read/write at the same time, while still allowing for one to
take over from the other.  The most obvious solution is some kind of
remote controlled power-off solution.  The failover should attempt to
shutdown the master and then turn-off power to the master before
taking over the disks.

Another thing to consider is what kind of filesystem to use.  To avoid
corruption of data and long fscks during failover, you will want to use
a journaling filesystem such as ReiserFS, Ext3 or perhaps xfs (SGI) or
jfs (IBM).

> Hello list members,
> We are about to set up a solution using piranha running under RedHat 
> Linux 6.2 to enable fail over services for a file server "cluster" with 
> samba as main application. Before we commit totally, it would be 
> interesting to hear if somebody has tried this before (we like 
> inventing wheels, but only if no round ones are to be found ...). In 
> the docs we have trawled the Internet for, we have not seen anything 
> about this particular combination. Any comments would be appreciated.
> Best regards and merry Christmas,
> Ragnar Wisløff
> LinuxLabs AS

John Cronin

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]