hpasm seems to freeze rhel 3

[Prud'homme Sébastien]

Hi,

I've installed HP management agents on 100 ProLiant ML 370 G3. Since then i noticed that some servers sometimes freeze (ping ok, but rlogin not ok, no access by gdm). I have to reboot them "with the button"...

Here is the configuration :

* ML 370 G3
* ROM Version 09/08/2003 Family 386P28 Type 03

* Red Hat Enterprise Linux 3 Update 1
* kernel-smp-2.4.21-9.EL
* glibc-2.3.2-95.6
* libstdc++-3.2.3-24
* compat-libstdc++-7.3-2.96.123

* hpasm-7.1.0-145
* cmanic-7.1.0-5
* cmastor-7.1.0-12
* ucd-snmp-4.2.5-79 (from HP website)

FYI, I'm NOT using drivers from HP website (for nic NC 7781, storage controller Smart Array 641)

[Steven E. Protter]

I'm not sure you can use this product with non-hp drivers.

I'm also not sure the hpasm is even certified on RH 3

SEP

[dirk dierickx]

AFAIK you don't need any hp driver/software to get linux running on their hardware. it runs just 100% fine without.
we have a boatload of hp servers here, and _never_ add the hp software, just RH as-is and we never had a single problem so far.
i just don't like things messing with the kernel.

[Prud'homme Sébastien]

I really need to use hpasm on these servers cause it's the only way to say when there are hardware problems (raid disk failures, fan problems, cpu crash)

[Don_89 · Accepted solution]

Uninstall the 7.0 version of HPASM drivers and use version 6.40. These are alot more stable. We also use only Red Hat drivers.

This article can also help you determine the problem..

http://www.linuxdevcenter.com/pub/a/linux/2001/11/01/postmortem.html

[Edmund White]

I bet it's the firmware on your Smart Array 641 controller....

I recently experienced a problem with the SA641 controller on ML350 and ML370 servers that caused the system load to rise very rapidly (> 40), halting most network services. It appeared as though the controller would shutdown and that processes that depended upon Disk I/O would go into STAT D (uninterruptible sleep), forcing the load up by one unit per process. Programs loaded into memory (the kernel, top, etc.) were unaffected. This always occured after 3-7 days of uptime (usually when physical memory was cached and swapping occurred).

This problem was fixed by replacing the 641 with a 6400 or 5300 series controller... OR downgrading the firmware (to the last revision from 2003) on the 641. The new firmware on the 641 was just released last week, and seems to have corrected the issue.

I spoke with several HP techs, as I have about 100 systems around to country to support. They told me to simply stop selling that raid controller until they released a new firmware. Messy. All of my systems are RedHat 8.0, run the 6.40 hpasm and cmastor drivers and use 5300, 6400 or 64x series raid controllers with custom vanilla 2.4.21 or 2.4.26 kernels. I experienced this in a repeatable fashion on a new ML350, but a coworker had the same issues with RHEL 3.0 on a 641-equipped ML370 and the 7.0 agents.

The bad firmware is the March 2004 Smart Array 1.92A. The good ones seem to be 2.26B or 1.30.

http://h18000.www1.hp.com/support/files/server/us/download/21214.html

So in this case, try downloading the new firmware for the 640 and see if it stops the crashes. To test, you may want to leave a console running top open on the server and watch the load rise. Most services will stop responding after the load hits 40+.

[Prud'homme Sébastien]

Thanks for your answer. Unfortunately, the Smart Array 641 firmware was not updated (still 1.30B). I will try to update to the last version of system ROM as it seems it could correct the problem with hpasm