Hide Forgot
Created attachment 526044 [details] lspci -vvv Hi, Since I upgraded to kernel-xen 2.6.18-274 and 2.6.18-274.3.1 I experience hard system resets on this server. This is Dell PowerEdge 1950 running 5.7 up2date and and is running XEN with 5 VMs. Logs do not show anything. No hung, no reboot. It's like a reset every 1 or two days without any reason. No special load on the host or the vms. With previous kernel 2.6.18-238.19.1 I don't have this kind of problem. # last | grep reboot reboot system boot 2.6.18-274.3.1.e Mon Oct 3 12:26 (02:10) reboot system boot 2.6.18-274.3.1.e Sat Oct 1 22:49 (1+15:47) reboot system boot 2.6.18-274.3.1.e Sat Oct 1 19:47 (1+18:49) reboot system boot 2.6.18-274.3.1.e Sat Oct 1 15:30 (1+23:06) reboot system boot 2.6.18-274.3.1.e Sat Oct 1 01:30 (2+13:07) reboot system boot 2.6.18-274.3.1.e Fri Sep 30 19:48 (2+18:48) ... reboot system boot 2.6.18-274.el5xe Mon Sep 26 12:45 (00:13) reboot system boot 2.6.18-274.el5xe Sun Sep 25 19:58 (17:00) reboot system boot 2.6.18-274.el5xe Sun Sep 25 15:40 (21:17) reboot system boot 2.6.18-274.el5xe Sat Sep 24 14:33 (1+22:25) ... reboot system boot 2.6.18-238.19.1. Sun Aug 21 15:20 (24+00:24) reboot system boot 2.6.18-238.12.1. Wed Jul 20 14:20 (32+00:57) Any help on debugging this? System's hardware seems ok at least from Dell management software. best regards, Giannis
Try running the latest kernel (-286) on it to see if it still happens. Also set it up with crashdump to get a core next time it fails (see the instructions below how to do that). You can also poke through your /var/log/messages now to see if there's any clues. Look for similar logs that popped up at or before the reboots. You can set up the host to capture a dump as follows 1) set crashkernel=128M@32M on the xen.gz line (256M instead of 128M may be necessary) 2) Make sure kexec-tools is installed 3) Make sure the bare-metal kernel is installed (it will kexec into the bare-metal kernel) 4) 'service kdump start' and/or turn it on for runlevels needed You can test that it works by triggering the crash in some way, such as "echo c > /proc/sysrq-trigger" or by using ctrl-a ctrl-a ctrl-a on the console for Xen's debug prompt, then the 'C' command to generate the dump The dump will be in /var/crash/<date>/vmcore (check that there's enough disk space for it before)
Hello Giannis, did you have any luck with capturing a dump (comment 1)? Thanks.
Hi Giannis, if you have the kernel dump, please reopen, and also state whether you have a RHEL subscription. Thank you very much.