March 27, 20179 yr Guys, I have 2 VMs running using the default settings for Fedora 25 Server/Ubuntu 16.04 (running the latest upgrades) with 1GB of RAM on a 6GB Core i7 920. Server is not under load, yet I see these troublesome error messages on the console from the very beginning on both systems: Any ideas? The domains are running on a dedicated Crucial 750GB SSD exclusively for these VMs. Any ideas? The Intel CPU is cooled with a Corsair H80i. So heat or anything like that is not an issue. Message from syslogd@cruise at Mar 26 23:02:56 ... kernel:Dazed and confused, but trying to continue Message from syslogd@cruise at Mar 26 23:02:56 ... kernel:Uhhuh. NMI received for unknown reason 21 on CPU 0. Message from syslogd@cruise at Mar 26 23:02:56 ... kernel:Do you have a strange power saving mode enabled? Message from syslogd@cruise at Mar 26 23:02:56 ... kernel:Dazed and confused, but trying to continue Message from syslogd@cruise at Mar 26 23:03:26 ... kernel:Uhhuh. NMI received for unknown reason 31 on CPU 0. Message from syslogd@cruise at Mar 26 23:03:26 ... kernel:Do you have a strange power saving mode enabled? [ 402.003889] perf: interrupt took too long (2632 > 2500), lowering kernel.perf_event_max_sample_rate to 75000 [ 405.771804] perf: interrupt took too long (3450 > 3290), lowering kernel.perf_event_max_sample_rate to 57000 [ 439.125760] perf: interrupt took too long (4323 > 4312), lowering kernel.perf_event_max_sample_rate to 46000 [ 444.931888] perf: interrupt took too long (5416 > 5403), lowering kernel.perf_event_max_sample_rate to 36000 [ 452.202561] perf: interrupt took too long (6773 > 6770), lowering kernel.perf_event_max_sample_rate to 29000 [ 709.348689] perf: interrupt took too long (8659 > 8466), lowering kernel.perf_event_max_sample_rate to 23000 [ 1009.354226] perf: interrupt took too long (11027 > 10823), lowering kernel.perf_event_max_sample_rate to 18000
April 20, 20179 yr I'm getting this same issue. I wrongly assumed that I had bad setting and was "sharing" a CPU or had not mapped a CPU and it's threaded partner, but neither is true; Each of my VMs has distinct CPU assignments and the pairs are correctly matched. I have 3 VMs, two of which are currently shut down and the 3rd is still giving this message. I do have a container running that's chewing up a lot of CPU on the unRAID server. Could that be the problem?
April 20, 20179 yr Author Hi rbroberts, your Docker should not be causing this, as a test you can stop the containers. In my case I have the same VM issues with the container, and without. Of interest is the fact that it is Fedora causing this at the moment. Windows XP seems to lose sync with the mouse, but I am not sure if that is related to this. As a separate data point, once thing I have not tried is running Fedora using libvirtd under a separate Ubuntu instance. In unRAID, Ubuntu runs fine without any perf or NMI issues. Which VMs are you running? Can you provide some logging? My server has since been upgraded to a 6 core Xeon -- albeit the issues remain. Edited April 20, 20179 yr by daze
April 21, 20179 yr I'm running three VMs, Fedora 25 Astronomy spin, the one with the NMI; Scientific Linux SL7 (based on RHEL 7), which wasn't giving NMIs but after I added passthrough for my NVIDIA 1050 card, it does. I removed the mapping and it's fine again. And lastly Ubuntu 16.04LTS, which has not had any issues. All of them have 2 real CPUs and the 2 associated threads assigned, all have 4-8 GB of memory and a 32GB vdisk, all are running SeaBIOS. The server is an Intel S2600CP2 with 2 Xeon E5-2695 v2 (12-core), 128GB or RAM, 6 x 4 TB drives. I had planned on passing through the NVIDIA card to be able to do CUDA programming on it, obviously not under RedHat relatives :-/
April 22, 20179 yr Author Wow. Nice to have a server. I seem to be growing my POS old PC into one... I am going to install Fedora 25 under Ubuntu in nested ESXi using libvirt, and on unRAID's libvirt if it'll let me know. Report back soon!
April 22, 20179 yr It's nice, or would be if the stupid VMs would behave. The point was to replace a pile of older machines with one not-quite-so-old (second-hand is very affordable) Right now, I'm not getting any NMIs, but the VMs are not really doing anything. (Edited because I left out that one little word "not") Edited April 22, 20179 yr by rbroberts
April 22, 20179 yr Author rbroberts, I have a NUC running ESXi. Everything works perfectly in it. It's low power and small form factor with a 512 SSD and 1GB RAM on a Core i5. For me too, would have been nice to consolidate stuff on this Xeon (already had ordered extra RAM for it).
Archived
This topic is now archived and is closed to further replies.