daniellee343
New Member
This is the second time I crashed our department servers. :-( Really don't know why.
The previous configuration was: We had a persistent memory namespace (/dev/pmem0) that was mounted at /mnt/pmem. We wanted to change it to Mixed Mode. So we unmounted this file directory ($ umount /mnt/pmem), disabled namespace ($ ipmctl disable namespace0.0), destroyed namespace ($ipmctl destroy namespace0.0), changed it to mixed mode ($ ipmctl create -goal MemoryMode=20 PersistentMemoryType=AppDirectNotInterleaved). All these steps were fine, until we reboot.
The reboot caused an infinite loop, according to the tech staff, and "the entire hard disk (/dev/sda) has been erased including the partition table." See attached. This is the part I got really confused about, since we did not do anything to /dev/sda. The /dev/sda was mounted under two places, "/", and "/var". So my guessing is either 1) since we did not clean up all the data under /mnt/pmem before we do umount, the data by default goes to "/" which is /dev/sda, causing some error I don't know, or 2) some other processes are busy with data under /mnt/pmem that we didn't use "lsof" to check and force to kill those processes.
I cannot recall any other command that caused this issue. Does someone have the same problem?
The previous configuration was: We had a persistent memory namespace (/dev/pmem0) that was mounted at /mnt/pmem. We wanted to change it to Mixed Mode. So we unmounted this file directory ($ umount /mnt/pmem), disabled namespace ($ ipmctl disable namespace0.0), destroyed namespace ($ipmctl destroy namespace0.0), changed it to mixed mode ($ ipmctl create -goal MemoryMode=20 PersistentMemoryType=AppDirectNotInterleaved). All these steps were fine, until we reboot.
The reboot caused an infinite loop, according to the tech staff, and "the entire hard disk (/dev/sda) has been erased including the partition table." See attached. This is the part I got really confused about, since we did not do anything to /dev/sda. The /dev/sda was mounted under two places, "/", and "/var". So my guessing is either 1) since we did not clean up all the data under /mnt/pmem before we do umount, the data by default goes to "/" which is /dev/sda, causing some error I don't know, or 2) some other processes are busy with data under /mnt/pmem that we didn't use "lsof" to check and force to kill those processes.
I cannot recall any other command that caused this issue. Does someone have the same problem?