Hello All.
I've just started looking into Xen and have a test environment in place. I'm seeing an annoying problem that I thought worthy of a post.
Config:
I have 2 x HP DL585 servers each with 4 Dual core Opterons (non-vmx) and 16GB RAM configured as Xen servers. These run CentOS 5.1 with the latest updates applied. These system both attach to an iSCSI target which is an HP DL385 running ietd and serving SAN based storage.
I have a test VM running CentOS 5.1 also updated.
Problem:
If I run the VM on a single server everything is OK. If I do a migrate of the VM to the other server I start getting random "BUG: soft lockup detected on CPU#?" messages on the VM console. The messages seem to happen with IO but not every time. A reboot of the VM on the new server will stop these messages.
I've also left the VM running overnight a couple of times and when I do I find that any external sessions (ssh) are hung in the morning but the console session is not. New ssh sessions can be started and seem to work.
After much googling it looks like the kernel messages can occur if dom0 is very busy but mine is not.
Any suggestions?
Regards
Brett Worth
My un-authoritative answer: I've been tracking this bug (or several with the same symptoms) for going on a couple years. It's ridiculously common, apparently well known to the Xen/Xensource guys judging by the number of reports/bugs posted, but I haven't seen mention of it actually being addressed and resolved. Unfortunately I see the same issue with it cropping up after VM moves, though it occurs /every/ time there is a VM migration, once per processor in the VM; doesn't matter if there is any IO on the Dom0 or DomU. Occasionally VM's die during a migration and have to be manually destroyed/restarted.
I do see evidence of significant instability (not implying it is related to the above softlockup issues) however, in either VM moves migrating from a Xeon (5345) to Opteron Dom0, and in high-utilization DomU's which are just plain flaky and reboot/die semi-frequently even when never altered from their start Dom0.
For me, it currently means running only low-priority non-production services in a VM, and not shelling out for RHEL5 support for the project (contrary to what I planned) since it's not being addressed. I'd be curious if this is being addressed in the Xen 3.2 release for RHEL5*...
Cheers,
/eli
Brett Worth wrote:
Hello All.
I've just started looking into Xen and have a test environment in place. I'm seeing an annoying problem that I thought worthy of a post.
Config:
I have 2 x HP DL585 servers each with 4 Dual core Opterons (non-vmx) and 16GB RAM configured as Xen servers. These run CentOS 5.1 with the latest updates applied. These system both attach to an iSCSI target which is an HP DL385 running ietd and serving SAN based storage.
I have a test VM running CentOS 5.1 also updated.
Problem:
If I run the VM on a single server everything is OK. If I do a migrate of the VM to the other server I start getting random "BUG: soft lockup detected on CPU#?" messages on the VM console. The messages seem to happen with IO but not every time. A reboot of the VM on the new server will stop these messages.
I've also left the VM running overnight a couple of times and when I do I find that any external sessions (ssh) are hung in the morning but the console session is not. New ssh sessions can be started and seem to work.
After much googling it looks like the kernel messages can occur if dom0 is very busy but mine is not.
Any suggestions?
Regards
Brett Worth
CentOS-virt mailing list CentOS-virt@centos.org http://lists.centos.org/mailman/listinfo/centos-virt
Brett Worth wrote:
Hello All.
I've just started looking into Xen and have a test environment in place. I'm seeing an annoying problem that I thought worthy of a post.
Config:
I have 2 x HP DL585 servers each with 4 Dual core Opterons (non-vmx) and 16GB RAM configured as Xen servers. These run CentOS 5.1 with the latest updates applied. These system both attach to an iSCSI target which is an HP DL385 running ietd and serving SAN based storage.
I have a test VM running CentOS 5.1 also updated.
Problem:
If I run the VM on a single server everything is OK. If I do a migrate of the VM to the other server I start getting random "BUG: soft lockup detected on CPU#?" messages on the VM console. The messages seem to happen with IO but not every time. A reboot of the VM on the new server will stop these messages.
I've also left the VM running overnight a couple of times and when I do I find that any external sessions (ssh) are hung in the morning but the console session is not. New ssh sessions can be started and seem to work.
After much googling it looks like the kernel messages can occur if dom0 is very busy but mine is not.
Any suggestions?
The soft lockup is technically not a BUG.
You will see these errors if an IRQ takes more then 10 seconds to respond.
In your case I would take a look at your iSCSI setup and the time it takes to migrate the VM from one node to another along with SCSI reserve/release setup on the iSCSI target.
I also have been using the Xen 3.2 RPMs off xen.org to CentOS 5.1 which good results, the VM migration may run smoother and quicker in Xen 3.2, but in doing so you take Xen off the reservation, if your OK with that it may fix your issues.
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
Ross S. W. Walker wrote:
Brett Worth wrote:
Hello All.
I've just started looking into Xen and have a test environment in place. I'm seeing an annoying problem that I thought worthy of a post.
Config:
I have 2 x HP DL585 servers each with 4 Dual core Opterons (non-vmx) and 16GB RAM configured as Xen servers. These run CentOS 5.1 with the latest updates applied. These system both attach to an iSCSI target which is an HP DL385 running ietd and serving SAN based storage.
I have a test VM running CentOS 5.1 also updated.
Problem:
If I run the VM on a single server everything is OK. If I do a migrate of the VM to the other server I start getting random "BUG: soft lockup detected on CPU#?" messages on the VM console. The messages seem to happen with IO but not every time. A reboot of the VM on the new server will stop these messages.
I've also left the VM running overnight a couple of times and when I do I find that any external sessions (ssh) are hung in the morning but the console session is not. New ssh sessions can be started and seem to work.
After much googling it looks like the kernel messages can occur if dom0 is very busy but mine is not.
Any suggestions?
The soft lockup is technically not a BUG.
You will see these errors if an IRQ takes more then 10 seconds to respond.
In your case I would take a look at your iSCSI setup and the time it takes to migrate the VM from one node to another along with SCSI reserve/release setup on the iSCSI target.
I also have been using the Xen 3.2 RPMs off xen.org to CentOS 5.1 which good results, the VM migration may run smoother and quicker in Xen 3.2, but in doing so you take Xen off the reservation, if your OK with that it may fix your issues.
After seeing this same issue on my Xen 3.2 install, but with NO migration or iSCSI happening I decided it is probably NOT iSCSI's fault, so I decided to research it a little more and this is what I found:
http://docs.xensource.com/XenServer/4.0.1/guest/ch04s08.html#rhel5_limitatio...
XenSource does provide a repo of CentOS 5 kernels that have been patched to fix this though:
http://updates.xensource.com/XenServer/4.0.1/centos5x/
But these seem to be woefully out of date.
I wonder if a kind soul would add the fix to the centosplus kernel with XenSource's patch so those rogue Xen users could benefit from this fix until upstream decides to include it.
I suppose the centosplus patch would need to be flagged interm in case it needs removed when upstream has their own fix.
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
Ross S. W. Walker wrote:
After seeing this same issue on my Xen 3.2 install, but with NO migration or iSCSI happening I decided it is probably NOT iSCSI's fault, so I decided to research it a little more and this is what I found:
http://docs.xensource.com/XenServer/4.0.1/guest/ch04s08.html#rhel5_limitatio...
Thanks Ross. I'll do more research into this when I get back to work.
The Release Notes say:
"Soft lockup messages after suspend/resume or live migration (Red Hat Bugzilla 250994 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=250994). These messages are harmless, but there may be a period of inactivity in the guest during live migration as a result of the lockup."
This seems to imply that the delay during the live migration is causing the messages. In my case the messages persist forever long after the migration. I might take a look at the patched kernel and see it there's a specific fix that could be applied to a more recent kernel.
Brett
-
Brett Worth -
Eli Stair -
Ross S. W. Walker