1

We have several 2012 Mac minis in our server room that reliably disconnect from 1GbE network (but otherwise appear to still be working) when the adjacent servers are under heavy CPU load.

We suspect RFI / EMI but can't be 100% sure as we don't yet have a good way to measure it and isolate the source. Other servers (over 50, all kinds including a bunch of xserves) in the same room are fine. Taken out of the server room, Mac Minis are fine.

The weird parts:

  • the disconnects started happening after we added three new servers (render nodes) to an existing cluster of three, pretty much pinpointing them as the source of the issue
  • no issues when those servers are idle
  • issue seems to start when all six are under 50% or higher CPU load
  • all other devices in the server room - lots of them, all kinds - have no issues; only the Mac minis are affected

What would you recommend to troubleshoot and address the issue? (Outside of powering down those new render nodes - which we do need and which don't seem to affect anything else.)

What we've tried:

  • put in a known good 2012 Mac mini and observe the same issue (we don't have Mac minis others than 2012)
  • move Mac minis around the server room; location within the room doesn't seem to matter
  • collect CPU utilization data from render nodes and other devices; response times and timeouts from Mac minis and other network devices using Solarwinds NPM; observe strong correlation between CPU load spikes on render nodes and network issues on Mac minis

P.S. What we haven't done yet:

  • to check for a possibility that it's just one server misbehaving, disconnect servers one at a time while putting remaining ones under 100% CPU load
  • research RFI / EMI measurement and isolation equipment and services
  • attempt to ground the Mac minis (their power cables are 2-prong, aren't grounded) or otherwise shield them from RFI / EMI

We do need the Mac minis as we're running out of Xserves, and certain apps only run on Macs.

P.P.S. Apologies if this is off-topic and/or in the wrong forum.

Thanks for any ideas!

2
  • 1
    Did you ever solve this ? (just out of curiosity) Jan 1, 2016 at 23:41
  • Sorta - the culprit was a specific NIC on an HP G8 server that when under network load, would cause the Mac Minis to disconnect. Once we virtualized the server, repurposed the HP one for other things that wouldn't require high network utilization - the problem went away. In other words it wasn't EMI/RFI, it was something to do with Mac Minis being sensitive to very specific events on the network. We got quite a few other nodes on the network including Mac Pros of many flavors, XServes, iMacs, MBPs, servers of many flavors and various appliances - none had issues, just the Mac Minis. Jan 3, 2016 at 1:32

1 Answer 1

0

The culprit was a specific add-on 4-port NIC on an HP G8 server that when under network load, would cause the Mac Minis to disconnect. Once we virtualized the server, repurposed the HP one for other things that wouldn't require high network utilization - the problem went away. In other words it wasn't EMI/RFI, it was something to do with Mac Minis being sensitive to very specific events on the network. We got quite a few other nodes on the network including Mac Pros of many flavors, XServes, iMacs, MBPs, servers of many flavors and various appliances - none had issues, just the Mac Minis.

The other part of it is that the issue would only happen with Mac Minis in the same room (Data Center) - so perhaps the way those Mac Minis were connected to the network in that room (cables, switches) also played a role.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.