Search results

first glace it looks like a panic, but it was sending syslog messages to it.. this would go on all day if I let it. but in your case it was locked? It was a tough one to diagnose and a hard pill to swallow when I told my supervisor I needed to spend about $4k on the cluster but still not...

ahhhh my syslog was too big even for pastebin pro, this is the kind of noise I was getting until each cpu was stuck and the server died, one thing I remember now is I sill had a concole, but could not ssh into it but tg3 is involved Mar 16 20:42:36 proliant02 kernel: BUG: soft lockup -...

looking at some of my old posts how about "hung_task_timeout_secs" in the logs anywhere?

I did it without rebuiding, pulled all the broadcom nics out (including the built in 4 port card that HP put in) and put in new ( i used intel). then in /etc/udev/rules.d delete the file "70-persistent-net.rules" and reboot and let machine reconfig for new nics

me too, I understand your logic and frustration. It looks like my logs have turned over, but still looking. But I have a gut feeling that if you check your syslog ( I think, maybe dmesg) you might find some sort of irq issue on each core (going from memory) as it dies. And as you have...

still putting my bets on the tg3 driver I fought this problem for about 12 months, random panics till I got fed up put some money into some different nics. No issues since. I'll see if I can find some of my old logs, but as I recall in my case I had a 24 core setup. Each core would...

lsmod , can we see the results?

I had all kind of issues with my HP machines that came with Broadcom BCM5719 and BCM5720 nics using tg3 drivers Many as you describe , replaced them all with intel and has been rock solid for 90+ days now.

I have some 385p G8 units that gave me grief with random panics. They came factory with Broadcom chip set nics. Changed to intel and that all disappeared. YMMV But they have been stable now over 90 days+ Your timeline sounds like mine. Edit, dug thru some logs, it looks like for me it was...

Great news, I always would find a hole to fall into when setting up fencing and getting things all whacked.

Not sure what to add, but your graph shows one node. Do all nodes show same gap@time, or is it just affecting one node?

ah heck , same problem with either the 64 or 32 bit drivers.. anyway its outside of the scope of this thread.. But the QEMU guest agent will be a great help, thanks

updating existing install, everything else went ok. Just updating viostor drivers caused it to hang at boot.

Curious how anyone has done with the w2k3 64 bit viostor drivers? Would not boot, went to the w2k8 drivers and booted ok..

Interested to see the answers you get, I think this has been discussed before and I don't think cpu units applies to KVM, only VZ.

I finally did find the answer in B&W http://pve.proxmox.com/pipermail/pve-devel/2014-June/011699.html

Check /etc/apt/sources.list Make sure its on wheezy (not stable)... I had 1 of 4 servers do this. Sure it was my fault.

Was testing this option on a 2003 VM. Enabling more than one prevents the VM from starting. Wiki did not say much about it, outside of the description of the option. Red Hat gave a command # ethtool -L eth0 combined M Since this VM is on a bridged attached to a bonded set of interfaces...

nice.. thanks

Not a direct proxmox issue. Guessing with all the different spins out each one may handle this differently. I just updated a network driver, had to build it. Part of the process was needing the kernel headers for the build process. Driver built and installed and is working so far. My...