cciss driver notification

[Corey Shields]

Greetings,

With the old cpqarray driver, when a disk would go back in an array I would start to see kernel messages coming from that driver. With cciss, however, we've lost numerous drives and we don't know about it until we walk back to the machine room and see those evil lights.

Is there any way to get the cciss driver to throw up some kind of notice in the kernel log when something fishy comes up?

I've looked at the cmastor agents which would be great, but unfortunately they don't run under Gentoo Linux and I haven't been able to shoehorn it in yet.

TIA & Cheers!

-Corey

[Kodjo Agbenu]

Hello,

What version of cpqarray are you using ?
Another point, did you update the disk array controller with the latest firmware ?

Good luck.

Kodjo

[Corey Shields]

we don't use the cpqarray module anymore, that was my experience on an older system. Currently we are using the cciss module from the stock 2.6.0 kernel source (we have an internal 5i controller and an additional 6404 controllet). With 2.4.x, the 2 different controllers would seem to block each other. The 2.6 kernel fixed that issue. Anyway, that's a whole different story.

To answer your other question, we are up to the latest firmware as of about 3 weeks ago.

So the big issue is that when a drive goes bad, we have no way of knowing it (no notification).

Thanks!

[Edmund White]

Not really.... Unless you could get the cmastor or hpasm agents to work, I don't think there are any hooks in the cciss driver for what you'd need. Are you running Gentoo in production, or is this a test box?

[Corey Shields]

Yeah, this is a Gentoo production box. I work for a major research university, and we are putting Gentoo into our production environment more and more lately.

When I was trying 2.4.x on the box, I was able to get the hpasm kernel modules to work. The agents and rest of the package took some work because it came in a RedHat-specific RPM, but all of that changed when we went to 2.6 anyway. HP has told me that right now there are no plans to port the hpasm kernel modules to 2.6. I'm sure they would build, but the module build environment is too different in 2.6 and I really don't have the time to fuss with it (I'll buy a 6-pack for the guy who does, though). I think the program we need to query the drives is the "survey" program. This is distributed as a binary that depends on the hpasm kernel modules. I'm sure the binary would run under Gentoo with no problem, but where we are running a 2.6 kernel without those modules, survey doesn't do us much good.

All of this trouble lead me to ask about the module itself, because if the kernel module supports some kind of error when something goes wrong, we don't need the extra mess. One would think that some kind of checking would/could be done at the module level, but then again I am not a driver hacker. :)

Cheers!

[Edmund White]

I just scanned through the cciss.c source in the 2.4.21 kernel. I didn't see anything that would indicate that the driver could detect individual disk failures.... I think it's all in the survey module...

[Tim Small]

I had a quick hack at getting hpasm to compile under v2.6, but gave up after a while. If someone would like to keep it up, here is the diff:

http://buttersideup.com/files/hpasm-2.6-prelim.patch

I got a couple of modules compiled, but gave up when the cpqevt module seemed to refuse to load - despite having no remaining unresolved symbols...

Once installed, each module would need to be made like this:

make V=1 -C /usr/src/linux-2.6.5/ SUBDIRS=$PWD modules

[Tim Small]

BTW, just in case anyone from HP happens to be actually readint this...

Isn't this sort of thing your job? Based on this apparent level of support, do you think I'll recommend Compaq, or Rackable Systems hardware to my clients?