RedHat LINUX on Proliant - Supported or Not?!

[Mark Satayathum]

We have been battling ASRs, etc. for nearly 2 years; patch after patch, up2date after up2date, service request after service request, tar after tar (Oracle RAC/OCFS seems to be particularly fragile in this environment). The hardware platform doesn't seem to matter, DL360s, DL380s, DL580s (580s do seem to have more issues, but that is what we are running Oracle on). At one time we had weekly meetings with HP, RedHad, and Oracle technicians. Sure, everyone SAYS that they are partnered with everyone else, until it comes time to solve a problem. Below is the latest response from RedHat (names removed):

Last Update:
Mr. xxxxxxxx,

I have been looking over your sysreport, as well as your system logs.

One thing I noticed immediately looking over your system logs is that they are flooded with "unable to locate char-major-237" messages. This is literally happening several times per minute. But there is no character device defined by a stock Red Hat kernel with a major number of 237.

Your system has at least three non-Red Hat kernel modules loaded, and your kernel is tainted.

cpqci
cpqasm
cpqevt

These modules are Compaq/HP server management modules, which are not provided by Red Hat, and are unsupported. They are known to cause problems such as server crashes and lockups.

At this point, we must ask that you remove these kernel modules (and stop them from loading), and also stop the HP/Compaq server management software, and then see if you can still reproduce your problem. If you can still reproduce this problem, we will be glad to continue working through this with you to help solve the problem.

Thanks,
xxxxxxxxxxxxxxxxxxxxx
Red Hat Global Support

So - will there ever be a "Red Hat-supported" version of the management agents?
Or should we just accept the fact that the Red Hat/HP management agents time bomb is just going to go off every now and then?

Thanks for letting me vent (and yes, we are now evaluating different server vendors - if you can't stand the heat, get out of the kitchen).

[Steven E. Protter]

You want the management agent to work. HP does have a table of servers and Red Hat Versions that support that.

Your response center can help you find the table, I can't locate it today.

You will find the versions of red hat supported are somewhat limited. If you can live with that and the hardware is on the matrix, you can find a supported combination.

Lastly, you might consider dropping Red Hat support and getting an HP support contract that covers Linux. Their support policy and philosophy is much better than Red Hat's.

SEP

[Paul Cross_1]

I just wanted to second the opinion stated above that HP's linux support is definitely superior to RedHat's. I have paid for redhat support in the past with no benefit whatsoever.

[Mark Satayathum]

All of our configurations are supposedly supported and certified by HP and Red Hat. We were led to believe (by all vendors) that this combination would work when we first considered doing it.
It seems like whenever we have a problem, the solution given is to unload the management agents (Red Hat) or to "patch everything" (HP). Indeed, the two cannot agree on what NIC driver to use (HP says use the Broadcom driver, RH says to use the TG3 driver). On some of the servers, we purchased Intel NICs, just to get away from that argument. Pretty frustrating.

[HGN]

Hi

We also have the same kind of support issues, RedHat always want the kernel to be upgraded and there are no issues with the management agents as per HP, sometimes debugging becomes an issue for an hung state or crash, we enabled dumb to be captured on another server to investigate.

Rgds

HGN

[Ross Minkov]

I just want to add that buying Linux support from HP Services is not only better, but also cheaper.

And yes, you want these agents -- this is part of what differentiates Linux on HP ProLiant servers from Linux on some no name server.

Regards,
Ross

[Ross Minkov]

You can find the ProLiant server support matrix at the following URL:

http://h18004.www1.hp.com/products/servers/linux/hpLinuxcert.html

[Johannes Krackowizer_1]

hi mark,

i have heard of some issues where the hp agents crash the machines and there are two workarounds to this problem but both aren't what you propably want:

1) disable hp agents
2) disable smp (use only one cpu in a smp environment) and disable apm/acpi

i don't think that hp will close this bugs in their software because the problems aren't very young, they are known since a long time and nothing realy happens. sorry

best regards,

johannes

[Samantha Bacon]

Hi Mark
I can't help with the question but in my last company we were working on a voice-based system running on Linux/Proliants. We had continual problems with the Compaq NIC agents especially until Compaq gave up and told us to switch them off to stop the kernel panics. I understand your frustration!

[Brian Shepard]

Hi Mark,

I can relate to your problems. As one of the posts mentioned, it's a matter of getting the right combinations of kernel level, drivers, firmware, etc exactly correct. My Windows admins tell me it's a similar situation, but not as bad as Red Hat. It's a challange. One thing I can say is that the latest release of the management agents appears to have gotten rid of our problems. The policy where I work is that the HP management agents must be running on all our HP servers. I wanted to disable them permanently and have for a while. I have been testing the latest release (hpasm-7.1.1-87) for some time on 3 of our HP servers running Red Hat AS 2.1 and have not had problems. I am about to enable the new agents on all of our Red Hat servers very soon. I'll let you know how everything goes. I find that HP support for Linux issues is pretty good. I think RHN is very good if you have the Provisioning Module. Getting the Provisioning Module is extrememly important if you're running Red Hat on HP servers. It allows you to down level very easily. At any rate, I feel your pain and you're not alone if that's any consolation.

-Brian