Any guide to replace a xfs logdev?
Hi,
Any guide to replace a xfs logdev?
case 1: logdev device failed.
case 2: replace the logdev with a new NVDIMM-N device.
but I failed to find out some guide to replace a xfs logdev.
Best Regards
Wang Yugui ([email protected])
2022/02/08
Quoted 8 lines · Wang Yugui
Hi,
Any guide to replace a xfs logdev?
case 1: logdev device failed.
case 2: replace the logdev with a new NVDIMM-N device.but I failed to find out some guide to replace a xfs logdev.
The external log is specified at mount time on the mount command line, so all you should need to do is use the "-o logdev=/dev/XXX" option to point at the new device, which must be at least as large as the old device, and should be completely zeroed.
If the filesystem was cleanly unmounted, this should be all that is required.
If the log device failed during operation, the filesystem is probably not consistent, and you will most likely need to run an xfs_repair, zeroing out the log and repairing any inconsistencies, since your log device has failed and no longer replayable. So, something like:
# xfs_repair -L -l /dev/newlogdevice /dev/xfsdevice
where /dev/xfsdevice is your data device holding the filesystem, and /dev/newlogdevice is your new/replaced log device.
Before you actually do this, you might want to see if anyone corrects my statements or notices anything I missed. ;)
-Eric
Quoted 6 lines
Best Regards
Wang Yugui ([email protected])
2022/02/08
Hi,
Thanks a lot.
'xfs_repair -L' works well to replace a xfs logdev. # kernel: 4.18.0-348 / 5.15.21
but a qestion 'cleanly unmounted' below.
Quoted 8 lines · Wang Yugui
Quoted 7 lines
Hi,
Any guide to replace a xfs logdev?
case 1: logdev device failed.
case 2: replace the logdev with a new NVDIMM-N device.
Quoted 8 lines
Quoted 1 line
but I failed to find out some guide to replace a xfs logdev.
The external log is specified at mount time on the mount command line, so all you should need to do is use the "-o logdev=/dev/XXX" option to point at the new device, which must be at least as large as the old device, and should be completely zeroed.
If the filesystem was cleanly unmounted, this should be all that is required.
'mount -o logdev=/dev/newlogdevice' failed,
# kernel: 4.18.0-348(centos 8.5) / 5.15.21
the message in dmesg:
Corruption warning: Metadata has LSN (1:6) ahead of current LSN (1:0). Please unmount and run xfs_repair (>= v4.3) to resolve.
/dev/newlogdevice is with a same size, and zeroed by blkdiscard(SSD)
'umount' is not enough to get a ' cleanly unmounted'?
Quoted 5 lines
If the log device failed during operation, the filesystem is probably not consistent, and you will most likely need to run an
xfs_repair, zeroing out the log and repairing any inconsistencies, since your log device has failed and no longer replayable. So, something like:
Quoted 7 lines
# xfs_repair -L -l /dev/newlogdevice /dev/xfsdevicewhere
/dev/xfsdeviceis your data device holding the filesystem, and/dev/newlogdeviceis your new/replaced log device.Before you actually do this, you might want to see if anyone corrects my statements or notices anything I missed. ;)
Best Regards
Wang Yugui ([email protected])
2022/02/09