I'm attempting to figure out what's causing this on my home lab server. It's an old Dell CS24 (dual Xeon L5420 CPUs, Intel S45-based motherboard). It's a fresh installation. When sitting completely idle (no VMs running, no other operations pending/running), one core on one CPU will be running at 100%. This is also a stock build of ESXi. All the hardware on the server appears to be supported, so there's no custom drivers or anything like that. The one modification is that I add the "ignoreHeadless=TRUE" argument to the kernel. Same issue once I do start some VMs up. The core remains maxed out. Needless to say it's robbing me of some performance, and using more power than it should be.
Esxtop shows the clulprit as "system" (process ID 2), which doesn't really help narrow it down. It starts shortly after the server boots, and never stops. Any ideas?
Wow, this forum is a graveyard. I would have thought there would be more discussion given the recent release of ESXi 6.
Just an additional note. It seems the problem doesn't start until a minute or so after the host has finished booting.
ESXi6 was released, but HCL was not updated. And that's a problem: nobody wants to risk upgrading 5.5->6.0 just to find out something is not working/supported...
Concerning your question: what does esxtop show? You should see the process loading 1 core...
Esxtop shows the clulprit as "system" (process ID 2),
Yeah, at this point I'm probably going to go back to 5.5. It never exhibited this behavior. As stated, it's the system process. It does indeed only load one core; it picks one at random and maxes it out.
OMG, I was thinking of migrating to ESXi 6 but now I'll better wait till ESXi U2.
This is just too bizzare.
Hmm... well I just blew a major hole in my theory that it was ESXi 6's fault. I installed ESXi 5.5 Update 2, and the issue is back. Now I'm really stumped. I swear it wasn't doing this before...
---
All the way back to ESXi 5.1 Update 3, and the problem has gone away. I've been running 5.5 for a while... maybe I just didn't notice it before.
I'm running ESXI 6 and I have this same problem as well. It looks like it has to do with IPMI. You can confirm this with ps -T.
Apparently, this was supposed to be fixed in 5.5u1. See links below for more details.
I'm thinking I just didn't notice it before. The good news is that 5.1 has been running fine. It may not have all the latest features, but my lab can't use those features anyway.