Bumping my email here for your attention. Can anyone confirm that the below recovery in the case highlighted is reliable and correct. On Wed, Jun 5, 2024 at 4:33 PM Lakshmi Narasimhan Sundararajan <lsundararajan@xxxxxxxxxxxxxxx> wrote: > > Hi Team, > A very good day to you all. > > I have a below scenario wherein mdraid array (raid0 originally) is > failing to assemble after a node reboot, while array was expanding > capacity (hence raid4). > Can you please confirm that the forced recovery in this below scenario > is correct and reliable always. > > 1/ create md array (raid0) > example cmd: > mdadm -C /dev/md/vol0 -n 2 --metadata 1.2 -c 1024 -l 0 /dev/sda /dev/sdb > > 2/ array goes full, so need to expand capacity > Since this is a virtual environment, I am able to resize /dev/sda and > /dev/sdb disks outside that are attached to my server node. > > After increasing the disk capacity backing disks /dev/sda and > /dev/sdb, I am attempting to grow the array to the max capacity. > > 3/ Convert md array to raid 4 > mdadm -G /dev/md/vol0 -l 4 > > 4/ grow array to max capacity > mdadm -G /de/vmd/vol0 --size max > > 5/ put it back to raid 0 > mdadm -G /de/vmd/vol0 -l 0 > > In the above sequence, I only expand capacity by resizing the array > elements and not by adding new disk to the array. While in the above > sequence, I hit a power fail and node got rebooted. > > After node reboot, since array was in raid4 it failed to come online. > ``` > Jun 03 23:08:35 kernel: md/raid:md126: not clean -- starting > background reconstruction > Jun 03 23:08:35 kernel: md/raid:md126: device sda operational as raid disk 0 > Jun 03 23:08:35 kernel: md/raid:md126: device sdb operational as raid disk 1 > Jun 03 23:08:35 kernel: md/raid:md126: cannot start dirty degraded array. > Jun 03 23:08:35 md/raid:md126: failed to run raid set. > Jun 03 23:08:35 kernel: md: pers->run() failed ... > ``` > > I recovered the array using the below sequence, given the array > expansion was only through disk capacity resize. > > mdadm -S /dev/md126 /// which mapped to /dev/md/vol0 above > mdadm -C /dev/md/vol0 -n 2 --metadata 1.2 -c 1024 -l 0 --assume-clean > --force /dev/sda /dev/sdb > > This brought up the array back online and contents were okay too. > > So in the above sequence of actions, I want to understand the following: > -- is the above method a reliable way to recover array incase there > was a node reboot while in raid4 state? > > -- Please let me know your thoughts on recovery and data reliability > on the array while recovering through the above sequence. > > Many thanks and kind regards for your help and insights. > LN
