Xserve activity lights

[jpaquin]

We have a cluster of 10 xserves here, and some of them will give us what seems to be a code with the system activity lights on the front. The activity lights that typically show cpu activity no longer activity show processing going on, but instead will have every other light stuck on. So, lights 1, 3, 5, and 7 on the top and bottom rows will be lit up permanently. Reboots, zapping pram, etc… won’t resolve it. If a machine in the cluster crashed sometimes it will come up with the odd activity lights as well.

What is more interesting is that when this happens if you boot to another OS (i.e. my firewire drive or the install DVD) the lights will function properly. Then, reimaging fixed 2 of the serves that had the issue, while causing it to happen on another 2.

These are all quad-core intel xserves running Leopard 10.5.4, 4 GB Ram, base configuration without any services running. They are actually being used in a cluster for episode engine.

Thanks for any input you have,

Jason

[jpaquin]

If anyone is interested in the outcome, it turns out this was all due to different OS builds. The servers that displayed this light pattern at random had been built off of an OS build that was slightly older than the server itself. All of these servers shipped with 10.5.0, but apparently two of them were purchased 2 weeks before the others and their original builds were different. The client then mixed up all of the installation cds and rebuilt the servers at random… Some newer servers got the older discs. It all comes down to organization…

The interesting find is that the servers were functioning without any known issues (other than frequent crashes that were caused by the client installing the ram incorrectly) even when running off of the incorrect build. The only clues were the lights on the front.

[iain]

The quad core xserves that shipped with 10.4.x (10.5.?) had lights that represented each core’s activity, with 4 lights always lit up. When you upgrade the server to 10.5.x, the light behavior becomes like the original xserve’s. Perhaps this was what you were experiencing.

[jpaquin]

thanks, I know what you are talking about, but that would have only been 4 solid lights, 2 on each row, and we still would’ve seen activity represented on them. With an ‘on’ indicating a light on, and an ‘x’ indicating a light off the representation was:

oxoxoxox
oxoxoxox

These servers do have 8 cores, so it’s possible it was trying to represent that, but we still had no activity represented. We have at this point proven the light status to be the result of an OS build being installed that truly was never meant to go on that particular server.

Thanks,
Jason

[Author]

Posts