- Joined
- Nov 25, 2013
- Messages
- 7,776
Hi all,
I have always argued that bhyve was production ready and for the most part I still think that way. Now we see two things with bhyve acting "weird" and they may be related or not.
The reason to investigate and write this post: very rarely at the start of the day we will find one of our production VMs switched off for no apparent reason. I cannot find anything in the log files, be it middlewared.log, messages - or anything in /var/log/libvirt. Is there any other place where I should look for clues?
My suspicion is that this problem might be memory related, which leads to the next question. Why this?
If you look at the configured memory of all VMs, it's 96G in total:
And the host system has got a whopping 256G. Additionally I limited the ARC to 64G via sysctl:
So I have 86G free, I try to start a VM with 16G configured, and TN warns be I would be overcommitting memory? Why? We do not run any other service but VMs on these systems.
This is 12.0-U2.1 - the hardware configuration is the "production hypervisor system" in my signature.
I can upgrade to U3.1 in the next couple of days, probably even today after end of business.
Kind regards,
Patrick
P.S. Also related or possibly not: I *never* had any VM being powered off unexpectedly in my private environment, which is much tighter with the memory. Main difference: all my private systems are Intel Atom or Xeon, our hypervisor host @work is AMD Epyc.
P.P.S. It seems only Windows VMs are affected. And the Windows system event log shows an unclean shutdown/crash.
I have always argued that bhyve was production ready and for the most part I still think that way. Now we see two things with bhyve acting "weird" and they may be related or not.
The reason to investigate and write this post: very rarely at the start of the day we will find one of our production VMs switched off for no apparent reason. I cannot find anything in the log files, be it middlewared.log, messages - or anything in /var/log/libvirt. Is there any other place where I should look for clues?
My suspicion is that this problem might be memory related, which leads to the next question. Why this?
If you look at the configured memory of all VMs, it's 96G in total:
And the host system has got a whopping 256G. Additionally I limited the ARC to 64G via sysctl:
So I have 86G free, I try to start a VM with 16G configured, and TN warns be I would be overcommitting memory? Why? We do not run any other service but VMs on these systems.
This is 12.0-U2.1 - the hardware configuration is the "production hypervisor system" in my signature.
I can upgrade to U3.1 in the next couple of days, probably even today after end of business.
Kind regards,
Patrick
P.S. Also related or possibly not: I *never* had any VM being powered off unexpectedly in my private environment, which is much tighter with the memory. Main difference: all my private systems are Intel Atom or Xeon, our hypervisor host @work is AMD Epyc.
P.P.S. It seems only Windows VMs are affected. And the Windows system event log shows an unclean shutdown/crash.
Last edited:
