Savecore not working

sqrl · May 10, 2010

I am having trouble with some of my R710 servers rebooting.

I am using 7.1 AMD64

I discovered the first issue was hyperthreading so disabled that with machdep.hyperthreading_allowed and that made the system more stable but I am still experiencing unexplained reboots.

I have run the dell diagnostics for 12 hours to test the hardware and it found nothing.

I have configured savecore dumpdev="auto" and dumpdir=/var/crash

I have attempted to force a dump with savecore -vvf

Output

Code:

unable to open bounds file, using 0
checking for kernel dump on device /dev/mfid0s1b
mediasize = 4294967296
sectorsize = 512
magic mismatch on last dump header on /dev/mfid0s1b
savecore: unable to force dump - bad magic
savecore: no dumps found

Any ideas I am out of ideas, I see others have experienced this issue but have not found a solution online yet.

achix · May 10, 2010

Do you have adequate swap ?
What does [CMD=""]swapinfo[/CMD] show?
Swap should be large enough able to contain at least as much as your main memory.

In any case, a dumpon should be run to instruct the kernel of a candidate for a crash dump device.
do a man 8 dumpon.

additional things to check:

do you get this message during boot: "No suitable dump device was found." ?
is the link /dev/dumpdev present?

sqrl · May 10, 2010

Achix

Swapinfo shows the following

Code:

# swapinfo
Device          1K-blocks     Used    Avail Capacity
/dev/mfid0s1b     4194304        0  4194304     0%

My system has 8GB ram with very low load at the moment.

/dev/dumpdev

Code:

lrwxr-xr-x  1 root  wheel  13 May 11 07:12 dumpdev -> /dev/mfid0s1b

sqrl · May 10, 2010

Just checked man 8 dumpon as mentioned by achix

have discovered it will not dump to any device smaller than hw.physmem so I believe that is the reason for it.

will do some testing

sqrl · May 11, 2010

Okay after doing some tests with an appropriate sized swap I was receiving the error
with dumpon ioctl: operation not permitted.

An attempt to force a dump with savecore -f /var/crash /dev/md0 found the same error

Code:

unable to open bounds file, using 0
checking for kernel dump on device /dev/md0
mediasize = 9294967296
sectorsize = 512
magic mismatch on last dump header on /dev/md0
savecore: unable to force dump - bad magic
savecore: no dumps found

achix · May 11, 2010

sqrl said:
Okay after doing some tests with an appropriate sized swap I was receiving the error
with dumpon ioctl: operation not permitted.

An attempt to force a dump with savecore -f /var/crash /dev/md0 found the same error

Code:

unable to open bounds file, using 0 checking for kernel dump on device /dev/md0 mediasize = 9294967296 sectorsize = 512 magic mismatch on last dump header on /dev/md0 savecore: unable to force dump - bad magic savecore: no dumps found

Why do you need to setup a memory disk?

your problem is that you are trying to save a core which simply has not been created yet.
Increase your swap size to be double of your main mem.
Then wait patiently for the next panic, or write a program to call panic(9) by your self, or just type [CMD="reboot"]-d[/CMD] .