Description of problem:/var/log/messages shows failed to load kdump kernel after installation Version-Release number of selected component (if applicable): RHEL5-b1 weekly release 01 Nov 2006 How reproducible: Steps to Reproduce: 1.install RHEL5 Beta1 weekly update 01 Nov 06 2.cat /var/log/messages|grep fail 3.The error shown is kdump: kexec: failed to load kdump kernel kump: failed to startup Actual results:kdump: kexec: failed to load kdump kernel kump: failed to startup Expected results:no errors or fails Additional info:
Can you provide more information on 1) HW type that you saw the problem on 2) The exact version of the kernel that the kexec/kdump failed message 3) can you give RHEL5 latest snapshoot a try?
Created attachment 141148 [details] hardware deatils
Attachment from comment #2 should satisfy the NEEDINFO.
03:02.0 SCSI storage controller: Adaptec AHA-3960D / AIC-7899A U160/m (rev 01) 03:02.1 SCSI storage controller: Adaptec AHA-3960D / AIC-7899A U160/m (rev 01) on 2 Intel(R) Pentium(R) 4 CPU 2.80GHz with 1G memory
The problem here is that we install kexec-tools with the kdump initscript chkconfig'd on. In the case of a clean install, the user generally hasn't configured kdump yet. Given that we have the kdump firstboot module and system-config-kdump now, I'd say we flip the kdump script back to not being on by default, as both tools will take the necessary steps to turn it on, once its actually been configured (i.e., kernel boot params are set up to reserve kdump mem).
Per discussion w/nhorman, we'll set the kdump script to not be chkconfig'd on by default, and let the user enable it, be that manually, via the firstboot module or system-config-kdump. Fix will show up in build -140.el5.
From comment #6: > Fix will show up in build -140.el5. Sorry. I'm not clear which package are we talking about here?? This doesn't match either the initscripts or the system-config-kdump package versions.
Sorry, that'd be kexec-tools-1.101-140.el5 (or later) that carries the fix. (The kdump initscript is part of the kexec-tools package).
I regressed with kexec-tools-1.101-140.el5.x86_64 & kexec-tools-debuginfo-1.101-140.el5.x86_64 though it says kdump is operational, it again gave me the failed message says kexec: failed to load kdump kernel kump: failed to start up and no vmlinux file is getting created.
Can you clarify exactly what you're doing that is failing? Is this immediately following a clean install? The only change here is to not chkconfig kdump on by default, so that after a clean install, it doesn't try to start the first time the machine boots. Instead, you get the firstboot kdump config tool, which if you decide to configure kdump, will chkconfig kdump on, and reboot your machine, after which kdump should successfully start up.
I tried with RHEL5 B2 snapshot-3 where the kexec is kexec-tools-1.101-143.el5. It gives me the same error. this is failing even after "chkconfig kdump on" on every reboot. manual starting of service kdump start also failing
What do 'cat /proc/cmdline' and 'cat /proc/iomem' spit out? It sounds like crash kernel memory hasn't been reserved.
I seen this issue only on some systems like PE1850, PE6600 though I have 2GB of RAM on these systems. On some other systems kdump is working fine. the output of /proc/cmdline shows: ro root=/dev/VolGroup00/LogVol00 rhgb quiet
Created attachment 144267 [details] output of /proc/iomem
Okay, looks like you're getting errors logged because there isn't any memory being reserved for kdump. Your command line shows nothing along the lines of "crashkernel=128M@16M" and /proc/iomem doesn't show a reserved crash kernel region (should see something along the lines of "01000000-08ffffff : Crash kernel" in /proc/iomem). Installing kexec-tools doesn't automatically configure the crashkernel reservation, that part has to be done manually. Please try adding "crashkernel=128M@16M" to your kernel commmand line, either by editing grub.conf directly or via system-config-kdump, restart your system and see if that doesn't resolve the issue.
yes, it resolved the issue. now I am able to see crash kernel entry in /proc/iomem. This issue is fixed
Closing, based on comment #16.
(In reply to comment #17) > Closing, based on comment #16. hi,all now, I got the same message as Giribabu Bikki described above, the following are my configures: 1. /boot/grub/menu.lst title Fedora (2.6.33.1) root (hd0,0) kernel /boot/vmlinuz-2.6.33.1 ro root=UUID=2cab961e-aa22-447e-a574-fdfe4dea6754 LANG=en_US.UTF-8 SYSFONT=latarcyrheb-sun16 KEYBOARDTYPE=pc KEYTABLE=us rhgb crashkernel=128M@16M initrd /boot/initrd-2.6.33.1.img 2.when I compiled linux-2.6.33.1, the configures related to kexec and kdump are: CONFIG_SYSFS=y CONFIG_PROC_VMCORE=y CONFIG_KEXEC=y CONFIG_DEBUG_INFO=y CONFIG_CRASH_DUMP=y CONFIG_HIGHMEM4G=y CONFIG_SMP=n CONFIG_PHYSICAL_START=0x1000000 CONFIG_RELOCATABLE=y but my /proc/cmdline is "ro root=UUID=2cab961e-aa22-447e-a574-fdfe4dea6754 LANG=en_US.UTF-8 SYSFONT=latarcyrheb-sun16 KEYBOARDTYPE=pc KEYTABLE=us rhgb crashkernel=128M@16M",
hi,all I am so sorry for sending the last mail! now, I got the same message as Giribabu Bikki described above,and as follows: May 20 08:59:20 localhost kdump: No crashkernel parameter specified for running kernel May 20 08:59:20 localhost kdump: failed to start up the followings are information about kexec and kdump: 1. kexec-tools-2.0.0-32.fc12.i686 2. Intel(R) Core(TM)2 Duo CPU E7400 @ 2.80GHz with 2G memory 3. /boot/grub/menu.lst title Fedora (2.6.33.1) root (hd0,0) kernel /boot/vmlinuz-2.6.33.1 ro root=UUID=2cab961e-aa22-447e-a574-fdfe4dea6754 LANG=en_US.UTF-8 SYSFONT=latarcyrheb-sun16 KEYBOARDTYPE=pc KEYTABLE=us rhgb crashkernel=128M@16M initrd /boot/initrd-2.6.33.1.img 4. when I compiled linux-2.6.33.1, the configures related to kexec and kdump are: CONFIG_SYSFS=y CONFIG_PROC_VMCORE=y CONFIG_KEXEC=y CONFIG_DEBUG_INFO=y CONFIG_CRASH_DUMP=y CONFIG_HIGHMEM4G=y CONFIG_SMP=n CONFIG_PHYSICAL_START=0x1000000 CONFIG_RELOCATABLE=y 5. my /proc/cmdline is "ro root=UUID=2cab961e-aa22-447e-a574-fdfe4dea6754 LANG=en_US.UTF-8 SYSFONT=latarcyrheb-sun16 KEYBOARDTYPE=pc KEYTABLE=us rhgb crashkernel=128M@16M" 6. there is no "crash kernel" in /proc/iomem 7. before restarting,I executed the command "chkconfig kdump on" based on the following message, for resolving this question, what should I do? any advice will be appreciated! Best regards pengcz
Hi, Did you check /sys/kernel/kexec_crash_size? I think it's because the crashkernel reservation failed because you specified the position of crashkernel memory reserved. Here it's 16M, if reserving crashkernel memory at 16M failed, it will happen like what you are seeing. So could you just set crashkernel=128M in /boot/grub/menu.lst, and try it again. Thanks Baoquan