RHEL 4 + PE 1850 + EMC AX100 PowerPath
tomf at novia.net
Wed Jul 13 20:13:48 CDT 2005
BTW - With RHEL4 U1, I believe RedHat will be inlcuding the mutipath
updates that understand the Clarion service processor stuff.
On Mon, 2005-07-11 at 00:12 -0500, JACOB_LIBERMAN at Dell.com wrote:
> Im sorry to hear about the problems youre experiencing.
> If you need assistance connecting a RHEL3 server to an AX100, your can
> call the storage support group at 800.945.3355 and follow the
> appropriate prompts to the UNIX/Linux host attach team.
> Usually a host wont boot when the external storage is attached if the
> HBAs SCSI adapter aliases are listed before the internal boot HBA in
> /etc/modules.conf. Usually you can swap their numeric position and
> remake the initial RAMdisk to get past that problem. Please give us a
> ring if that doesn't work.
> In general, the GFS 6.x pool_mp service for load balancing and fault
> tolerance does not accommodate back-end trespasses because the AX100
> (and the whole Clariion line) are active-passive arrays, not
> active/active. However, I have seen GFS used in conjunction with
> PowerPath and it seems to work although I do not believe it is on the
> ESM at this time. Incidentally, powerpath also works in conjunction with
> OCFS if you choose to use it as a clustered filesystem instead of GFS.
> Running GFS on a single node (in an SLM configuration) also presents a
> single point of failure. It would be preferable to use the Redundant
> Lock Manager file locking mechanism if you have the hardware capacity.
> You can run the lock manager service on hosts that are not physically
> connected to the array as long as they have network access to the other
> lock management servers.
> There are a few Veritas VxVM solutions that are validated on the ESM,
> but I think its primarily validated under Solaris, and very version
> specific. I have talked to people who are using Veritas VxVM with DMP in
> conjunction with PowerPath on Solaris. I know a few of our customers use
> Veritas VxVM with DMP and PowerPath on Linux, but I do not believe that
> EMC has validated that configuration.
> Finally, if you are still in the pilot phase, I also want to suggest
> that an AX100 might not be the best array for a high volume oracle
> database since it uses ATA drives rather than fiber. Plus, ATA drives
> have only single backend loops, so you cant really load balance them
> across SPs effectively. The CX300, although more expensive, would give
> better performance with internal fiber drives.
> Thanks, Jacob
> > -----Original Message-----
> > From: linux-poweredge-bounces-Lists On Behalf Of Tom Ferony
> > Sent: Sunday, July 10, 2005 8:53 PM
> > To: linux-poweredge-Lists
> > Subject: RHEL 4 + PE 1850 + EMC AX100 PowerPath
> > I'm not sure what Dell recommends, but I know what works, and
> > I believe this is now certified by EMC. I happen to work at
> > a very large railroad that is piloting a EM64T Dell box
> > (running x86_64 RHEL 3) and have had horrible experiences so
> > far, but we have solved the lack of EMC support for Linux by
> > using RedHat's GFS on a single node to provide both failover
> > and round robin load balancing on RHEL 3. This solution, and
> > another involving veritas and dmp solution were not offered
> > by Dell and the funny thing is that Dell sent us the box with
> > 4 fiber channel adapters installed in the box (although one
> > was faulty, which we had to troubleshoot) for the purposes of
> > a pilot with a very high volume Oracle database. The
> > resolutions and problem determination came after a Dell
> > support person left the premises, I came back from vacation
> > and had most of it resolved within two days.
> > Now if we can get the pilot box to reboot while having the
> > fiber channel adapters plugged in and not panic when loading
> > the USB drivers, we may be happier.
> > We also had 4 Dell blades as pilots, two died. One of them
> > melted and a replacement board was sent in, but it was warped
> > so badly it couldn't even be installed. The blades were
> > installed in their own rack in a brand new data center that
> > is very cool and well ventilated.
> > We have 350 UNIX boxes and hundreds of Intel boxes, very few
> > Dell, but if someone convinces us that Dell can support Linux
> > and can convince us that the hardware problems we've had were
> > flukes, Dell might have the whole ball of wax, and we're
> > exploding with growth.
> > --
> > Tom Ferony
Tom Ferony <tomf at novia.net>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Linux-PowerEdge