On 11/10/19 9:14 AM, Alex Powell wrote:
Hi all, I had a disk fail on my BTRFS RAID5 array. It is still mounting but there are bad sectors which will switch the array to read only mode when used. I used "btrfs device delete /dev/sdd /mnt/data" to remove it from the array. However it seems that it is only partially removing it from the array and when it gets to the bad sectors it fails. localhost ~ # btrfs device delete /dev/sdd /mnt/data ERROR: error removing device '/dev/sdd': Input/output error
Could you please provide 'btrfs dev df <mnt>' we need to know the allocated chunk profiles. If its only RAID5, we continue to reconstruct the RAID5 for failed IO during the delete. But here its may be that there are IO errors from more than one device in the RAID5, which unfortunately is a fatal error. You could confirm the devices involved in the errors from the syslog and 'btrfs device stats <mnt>'.
What is best practice for removing the drive from the array in this situation?
Device replace is better as it will continue to fulfill the raid5 redundancy promises. Yeah but you need the spare device, which is a good practice in the data centers. HTH. Thanks, Anand
Kind Regards, Alex Powell
