We managed to reproduce using another machine this time and with less ram http://1drv.ms/1KO85AV
Same hardware though. -----Original Message----- From: "David Finster" <[email protected]> Sent: 31.05.2015 08:29 To: "[email protected]" <[email protected]> Subject: Re: [smartos-discuss] Windows KVM Crashing I think you’ll have to see if you can replicate the failure on a smaller KVM. Examining the core dump doesn’t yield much beyond: mdb: core file data for mapping at 102b000 not saved: Disc quota exceeded mdb: core file data for mapping at 100102b000 not saved: Error 0 mdb: core file data for mapping at fffffd7ff52b2000 not saved: Error 0 […] mdb: core file data for mapping at fffffd7fffddc000 not saved: Error 0 debugging core file of qemu-system-x86 (64-bit) from 95c0f004-d78d-11e4-8377-4ff514e3fcfc file: qemu-system-x86_64 initial argv: /smartdc/bin/qemu-system-x86_64 -m 65536 -name 95c0f004-d78d-11e4-8377-4ff514e3 threading model: raw lwps status: process terminated by SIGABRT (Abort), pid=8611 uid=0 code=-1 I’m guessing that the dumping ground for the cores ran out of disk space, which given the RAM assigned was 64GB sounds rather likely. Perhaps see if you can replicate with a smaller KVM, but otherwise perhaps anyone with some additional mdb/QEMU tips could jump in? On 30 May 2015, at 6:00 pm, [email protected] wrote: we tried to do everything by the book, but as for the lsi card i dont know.. i would guess its in JBOD mode since we dont use any hardware raid etc , just configured all disks using zfs into a mirrored pool (2x2 mirror, 1 spare, two ssd for write / read cache). i will check though. i hope the download speed is not caused by the source, since that is hosted at a provider, but it is a big tgz (didnt even now tar had an option E until then) … the kvms have rather a lot of ram assigned (up to 64GB) since thats the only thing running there really (the other 64 we wanted zfs to have) which would make the core dump rather large. we will try to reproduce on smaller kvms on monday. From: David Finster Sent: Saturday, May 30, 2015 9:50 AM To: [email protected] All seems like a rather standard configuration, aside from the LSI card which seems to be a Fusion card. Has that been taken back to IT mode or just running in JBOD mode? Probably not the source of the problem though. I’ve got a box downloading the dump file, but it’ll take 15 hours. On 30 May 2015, at 5:19 pm, [email protected] wrote: Here is the server info: https://gist.github.com/matthiasg/8eb40a64e803c6e6e2d0 Sorry Core Dump Download URL didn't copy correctly (this is a weird setup of a server): http://87.106.137.9/wwwroot/downloads/core.qemu.tgz Iscsi was still in use (that's where the data store is), but not by SQL server anymore. So there was less traffic via iscsi and it didn't crash under low load. I did have a number of similar crashes on a different server last year (also storage of data via iscsi), but I didn't think it was an issue since that server (test server) had extremely old drives on purpose and showed a number of data errors. The iscsi target is also hosted from that same server so its local traffic only. thanks From: David Finster Sent: Wednesday, May 27, 2015 7:15 AM To: [email protected] Is there anything under /zones/<vm uuid>/cores for the particular virtual machine? On 27 May 2015, at 3:02 pm, Matthias Goetzke <[email protected]> wrote: Vmadm start is indeed required. The vm is just off and needs to be restarted with vmadm start. What can I do do diagnose this issue ? Regards Matthias From: David Finster Sent: 27.05.2015 00:47 To: [email protected] Subject: Re: [smartos-discuss] Windows KVM Crashing Hi Matthias Without knowing more about the crash scenario, it might be difficult to diagnose. If the crash/BSOD originates from the virtio disk driver, then there may not be anywhere to write crash information to. It might be worth setting Windows to not restart on BSOD, which can be done through Control Panel -> System and Security (System if small icons) -> Advanced System Settings -> Startup and Recovery and uncheck ‘Automatically restart’. Unfortunately you would be sacrificing the auto-restart feature but when you next notice the box is down you should be able to jump on VNC and see what the error condition was. If Windows is just BSOD’ing and rebooting within the VM itself then the qemu processes won’t exit so the vm.log output is probably correct. The only time this would be different is if Windows isn’t actually crashing, but qemu is, but that would require a vmadm start <uuid> to get it going again. Thanks, Dave On 26 May 2015, at 9:36 pm, Matthias Götzke <[email protected]> wrote: Hi, we are currently experiencing the same issue: with SmartOS 20150219T102159Z. There is no apparent reason with about 100GB free in the pools and no entries in the windows system log other than it restarted for unknown reasons. the vm.log just shows the boot process, the previous logs are all empty. We did use virtio drivers from january and did switch back to stable from dezember in case that causes issues. But i am not sure how a driver failure could initiate a full reboot without leaving a bluescreen dump. did anybody solve this or track this down further ? thanks, Matthias On Wed, Jul 23, 2014 at 6:34 PM, Mark Creamer via smartos-discuss <[email protected]> wrote: Hi, I have a Windows 2008 R2 KVM that crashes constantly. There is nothing in the system log, except after it comes up there is the usual "previous shutdown was unexpected." There is no bluescreen (bugcheck) recorded. None of my other KVMs are having the same issue, and this one is a problem on both the original host I built it on, and another host I moved it to. While I'm troubleshooting further, is there an updated driver for the virtual disk? I'm not sure where to locate that and I thought I might try that next. Thanks -- Mark smartos-discuss | Archives <img tabindex="-1" src="https://www.listbox.com/images/feed-icon-10x10.jpg" border="0" data-ms-imgsrc="https://www.listbox.com/images/feed-icon-10x10.jpg0d558cf.jpg?uri=aHR0cHM6Ly93d3cubGlzdGJveC5jb20vaW1hZ2 [The entire original message is not included.] ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
