XServe G5 Windfarm issue

[imrazor]

imrazor,

I'm actually back in 10.04. I've been up for a little over ten hours today and it 'seems' like things are the way they used to be. I know that's not very concrete but...

I reset the motherboard (PMU) last night and shutdown/restart commands now function the way I'm used to.

I'm working on getting everything set up so I can test the server via JMeter to make sure it's going to stand up to the load I'm scheduled to throw at it this coming Friday.

I've been running on a Mac Mini for over a year and have never had any problems but I have a test coming up that is going to be taken by almost 400 students in two waves of approximately 200 so I don't want to take any chances. I'm working on setting up my XServe (fully loaded with 8GB of RAM) with the LAMP on one drive, the data directory for Moodle (the system I'm using) on another, and the data directory for MySQL on a third. If I can manage to get everything set up, an the server doesn't suddenly shut down on me life will be grand.

I'll post later this evening to let you know if my server made it past the 12 hour record

And again on Tuesday after hitting it hard with JMeter.

Jason

After yesterday's Karmic failure, I decided to zag instead of zig. I installed Debian Lenny on my G5. I didn't get any FATAL windfarm errors, though I did get some messages about unreferenced symbols. After spending a few hours configuring it, I managed to do about 1-1/2 hours of hard transcoding/compiling. No crashes so far, but more testing is needed.

What is JMeter? Could I use it to stress test my G5?

[kokoroexchange]

Damn! Well I guess I was being overly positive!!!

I went out for a walk with my wife and daughter and just got back and the server was off!!! Just a little shy of 12 hours this time...

This won't do!!!

I guess I'll try and install 7.04 and let it run all day tomorrow and see what happens.... This is really frustrating....

Jason

[imrazor]

Well, poo. Was the Xserve under any sort of load, or was it just idling? Before I left this morning I started Lenny on a transcode run that should last ~1 hour. Oddly, under Lenny ffmpeg doesn't seem to peg both CPUs at 100%. Instead it's about 60% for each CPU. More research is needed on this I guess. When I compiled ffmpeg on Lenny I had to leave out a couple of libraries, so that may be the difference. I may do some research and figure out how to get those two missing libraries on Debian.

When I get home from work, I'll check the G5 and see if it's still up. I may even hear from my fiancee if it survived the transcode run.

[kokoroexchange]

No load, just idling behind my router without even any access from outside my LAN.

I've reverted back to 7.04 and can already tell that the fans are running differently. Much quieter...

The problem is that since 7.04 is no longer supported, it's going to be a royal pain to install packages that I need (libnet-ssleay-perl openssl libauthen-pam-perl libpam-runtime libio-pty-perl libmd5-perl). I guess I'm going to have to fish around for them one at a time, download each, and then install them. Yuk!

Or maybe it would be a good idea to try something that is still supported....which would be....8.04...I guess. No guarantee that it's going to work the way 7.04 did for me though...

What a dilemma, spend the time getting 7.04 in the state that I want/need or give 8.04 a shot....?

Jason

[imrazor]

No load, just idling behind my router without even any access from outside my LAN.

I've reverted back to 7.04 and can already tell that the fans are running differently. Much quieter...

The problem is that since 7.04 is no longer supported, it's going to be a royal pain to install packages that I need (libnet-ssleay-perl openssl libauthen-pam-perl libpam-runtime libio-pty-perl libmd5-perl). I guess I'm going to have to fish around for them one at a time, download each, and then install them. Yuk!

Or maybe it would be a good idea to try something that is still supported....which would be....8.04...I guess. No guarantee that it's going to work the way 7.04 did for me though...

What a dilemma, spend the time getting 7.04 in the state that I want/need or give 8.04 a shot....?

Jason

I found a couple of suggestions for windfarm issues on a Debian PowerPC mailing list. The first is to do as follows with root permissions:

# modprobe windfarm_core

The other suggestion was to add a "windfarm_core" line to /etc/modules (which is a similar solution, just applied at boot.) I won't be home for several hours, so if you get a chance to try either with 10.04 today, let me know if you have any success.

The OP on that thread was also having issues with random shutdowns, so hopefully that will resolve our issues.

You could preserve your new 7.04 install by imaging to an identically sized partition with gparted (I think), dd or a similar tool. I've imaged OS X many, many times but have rarely done so with Linux.

Good luck and thanks...

[kokoroexchange]

Sorry, I missed your question about JMeter.

Here is the link

http://jakarta.apache.org/jmeter/

To be honest, I've never used it but it was recommended by a person on a different forum so I thought I'd give it a try. Haven't gotten that far yet though so unfortunately, I can't tell you much about it.

I'm working on giving 8.04 a shot at the moment. Will let you know how it goes.

Jason

[kokoroexchange]

Ok,

Well it took some time but I got 7.04 back up and running with all the packages and dependencies I wanted and need. Whew!!

I tried 8.04 but the install froze midway through. Right at the partition tool so I gave up on that one.

I contemplated trying adding the trick you found on the Debian forums but couldn't afford to spend any more time on this and had a pretty good feeling I could get 7.04 back up and running so that's where I am.

After I complete the task that is immediately at hand I may spend some more time trying to get up to a newer version but as long as it's as stable as it was in the past, I'm happy with 7.04.

Too bad I had to go through all of this but it looks like I'm in business. Hope it's not too early to say that. I'll let it run for a while now and make sure I don't suffer from a sudden shutdown (fingers crossed). If I do I think it's definitely a hardware issue as I never had shutdown problems with 7.04 in the past.

Jason

[imrazor]

Latest update: Found this old bug report...

https://bugs.launchpad.net/ubuntu/+s...ux/+bug/241726

According to the bug report, my model of Powermac (7,3) doesn't have windfarm support. No mention is made of Xserves, unless they're based on one of the Powermac model ID's in the bug report. While those windfarm modules failed to load, one did load: windfarm_pid. On a Lucid PPC Live CD I was successful in removing the windfarm_pid module without the fans going to warp speed. It is possible, but somewhat unlikely, that the windfarm_pid is conflicting with the therm_p72 module, which is used by Powermac7,2 and Powermac7,3 (my model) for thermal control. I'm going to try reloading Ubuntu, then blacklisting all the windfarm modules and see if I still get crashes. Unlikely, but I don't feel like learning my way around Debian.

Eric

EDIT: Additional info; I did an rmmod of therm_pm72 and sure enough my fans ramped up within a couple of minutes. When I reloaded therm_pm72 they instantly shut off. Time for a reload of Ubuntu...

[imrazor]

I've reloaded Lucid and so far things seem promising. I just ran some heavy compiling and transcoding for 4-5 hours and no crash. I accomplished this by doing

sudo rmmod windfarm_pid

However, no matter what I try I have been unable to blacklist either windfarm_pid, or any of the other windfarm modules. They still give a FATAL error at boot, but don't seem to be crashing the system, at least so far. I've worked around this by adding

rmmod windfarm_pid

to rc.local. My fingers are crossed...

UPDATE: I left the G5 running overnight with

cat /dev/zero > /dev/null

running in two separate terminals, and an OpenGL screen saver running. Despite the CPU running at full load for 10-11 hours, it seems to be holding up OK. I may have it licked, but I'll continue stess testing it for another 2-3 days before reaching any conclusions.

[imrazor]

Since I added

rmmod windfarm_pid

to my /etc/rc.local I haven't seen a single random crash or shutdown. I kept my system on heavy load for about 18 hours without a shutdown. So far the results aren't conclusive, but it does seem as if there's a conflict between the therm_pm72 and windfarm_pid modules. Hopefully Jason or someone else will find the info useful, though it only seems to affect a few models of Mac.