On 09/23/2013 12:04 AM, Dave Gomboc wrote: >> Ok. I can't determine how the superblocks ended up the way they did, >> but the first two chunks appear to follow the proper patterns. >> >> I think you're best bet is to disconnect two of the drives, leaving one >> that identifies as "0" and one that identifies as "3". > > root@sysresccd /root % ls -l /dev/disk/by-id | grep Hitachi | grep -v part1 > lrwxrwxrwx 1 root root 9 Sep 22 20:17 > ata-Hitachi_HDS724040ALE640_PK1310PAG5ZY0J -> ../../sdb > lrwxrwxrwx 1 root root 9 Sep 22 20:17 > ata-Hitachi_HDS724040ALE640_PK1310PAG62REJ -> ../../sdc > lrwxrwxrwx 1 root root 9 Sep 22 20:17 > ata-Hitachi_HDS724040ALE640_PK1310PAG62T2J -> ../../sdd > lrwxrwxrwx 1 root root 9 Sep 22 20:17 > ata-Hitachi_HDS724040ALE640_PK1311PAG4W5TS -> ../../sda > root@sysresccd /root % cat /proc/mdstat > Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] > [raid4] [raid10] > md127 : inactive sdd1[3](S) sda1[0](S) > 3907021954 blocks super 1.2 > > unused devices: <none> > > Should I be disconnecting sdb and sdc, disconnecting sda and sdd, or > does it matter? Actually, from that report, just do "mdadm /dev/md127 --run". > I should reboot using the rescue disk before attempting the forced > assembly, not my boot drive, right? Yes, but only necessary if the above fails. And if there's a partial assembly, you might need to use "mdadm --stop". > Sorry if the answers to these questions seem obvious to you: I want to > make sure that I understand you exactly. I am moderately terrified at > the moment. You have duplicated the disks. You have all of the insurance possible. >> Then use "mdadm -Af /dev/mdX /dev/sdY1 /dev/sdZ1" >> >> The "-f" will force the assembly without regard to the event counts. >> Then you can take a backup. Finally you can add devices as "new" ones >> to rebuild back to full redundancy. (Fix your timeouts before >> attempting the latter.) > > When following up on your advice to search for those other terms, I > saw several examples where people specified 7 seconds to the disk > drive using that control program, and also read somewhere that while > Linux's software raid will wait, that Linux's scsi subsystem has a 30 > second timeout. So, 7 seconds sounds good? Most traditional enterprise drives power up with t=7 seconds. The SSDs I've used use t=4 seconds. Keep in mind that the setting is forgotten when the drive powers down. You need the commands in rc.local or your distro's equivalent. Phil -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
