6400R - PCI hotplug and fan speed control

[Remigiusz Marcinkiewicz]

I've got a used 6400R server, it works pretty well for such an old hardware, but there are two problems.

First, PCI hotplugging doesn't quite work under Windows 2003 SP1 (it does under Linux 2.6, with the standard kernel's Compaq PCI driver - LEDs, latches, buttons and of course powering the slots off/on all work). I had configured the server (after resetting the configuration with the SW1 switch on the I/O board) with SmartStart 4.80 CD for Windows 2000, checked all the provided options for sane values and installed Win2k3 SP1, then the SNMP service and the latest Proliant Support Pack from HP Drivers & Support page. There were no errors during the installation, the Device Manager lists two CPQPHP devices and there even is the "Safely Remove Hardware" menu, listing Compaq Array 3200 and HP NC3131 network adapter (both 64bit, came with the server). However, this is where the good ends. All the 32bit boards won't work - in fact, the slots won't even power up. The button does nothing and I can't find any software control for powering the slots *up* (there is one under Linux and it works, for 32bit boards too). As if that wasn't enough - they work if I just disable the hotplug driver and reboot, thus making all the slots permanently powered. Tested with a sound card and a USB controller (cheap 5-port by NEC), no problems at all. In short, no 32bit with the hotplug driver loaded. But still, even the 64bit boards make problems. I've tried taking the NIC out, pressed the button, waited until the LED and OS both confirmed that the slot is powered down, unlatched it and lifted the board... Kaboom, BSOD, something about cpqarray driver (how the hell? It's on a separate PCI bus!) and dumping memory, but I couldn't read it all, the server just rebooted after a few seconds. Again, this all works under Linux, with both 32bit and 64bit boards, so the hardware is fine. I don't have even a faintest idea for a single one earthly reason of such behaviour and, being an everyday Unix admin, no knowledge of Windows internals and diagnostic methodology, and so I'm there. Oh, and Windows 2000 SP4 did the same weird things, so that's not just the 2003 SP1.

The other problem applies universally - neither Windows nor Linux could reduce the fan speed, despite being able to tell me the temperatures (31, 30, 27 Celsius in a ~20C room, at least it does its job) and assure me that the fans are running on "Normal" speed. I wonder, given that the server sounds like a small rooftop AC unit now, how would they sound on "High" speed, like a Boeing 747 taking off on my desk maybe?... Anyway, the Insight Manager Agents, Server Agents and Foundation Agents, among otheres, are installed, I can even see that the cpqhlth.dll module is loaded by some application - but the fans are still making terrible noise.

Thanks for the help,
Remigiusz "Enleth" Marcinkiewicz

[David Claypool]

The health driver contains the logic that controls the fans. If it is reporting to the agents that the fan speed is normal, that's the state it has placed the fans in. Keep in mind that this is a server, designed for data center use, not a desk side PC. Block the air intakes for a bit and you'll see how fast (and loud) the fans get. Are the fans over-engineered? You bet. That's one of the reasons for legendary ProLiant reliability. The 6400R was retired from the market 12/31/2000.

[Remigiusz Marcinkiewicz]

Of course it is a server, however, even on this forum I've seen reports of Proliants slowing down to sane speeds in good thermal conditions after the drivers were installed. In my case, there isn't really any difference between the server booting up and working with the OS loaded, neither there is any after blocking the intakes, taking a fan out for a while, or even cooling the room down to 15C. There's just no effect. Maybe I'm unable to hear the difference, maybe there isn't one because it's blowing the air with full force all the time - the easiest way to check that would be probably to override the default control, but I feel that's not something that Compaq programmers might have put in their drivers. In the worst case, I won't heistate to get the RE tools and rip the driver apart to see how it's done, but it'd be nice to try a sane way first.

Regards,