> On Wed, Aug 5, 2015 at 11:59 AM, George Rapp <george.rapp@xxxxxxxxx> wrote: >> On Wed, Aug 5, 2015 at 12:17 AM, George Rapp <george.rapp@xxxxxxxxx> wrote: >>> Hello - >>> >>> Fedora 22 user (kernel 4.0.4-303.fc22.i686+PAE) using mdadm - v3.3.2 - >>> 21st August 2014. (Of course, I don't have a backup ... 8^) >>> >>> I had a healthy RAID 6 array, and was trying to grow it from 5 >>> partitions of size 1.8 TB to 6 partitions. >>> >>> # mdadm --add /dev/md6 /dev/sdi1 >>> >>> # mdadm --grow --raid-devices=6 --backup-file=/home/gwr/c/grow_md6.bak /dev/md6 >>> >>> The second command threw a bunch of SELinux errors (ah, thank you, >>> SELinux, for always being there to bite me in the ass when I don't >>> expect it ... 8^) about access to /home/gwr/c/grow_md6.bak. The >>> reshape operation sat for many minutes at 0% progress, according to >>> /proc/mdstat. However, the file /home/gwr/c/grow_md6.bak *was* >>> created; it's about 6MB. >> >> Additional data point: the backup file is all zeros, according to an >> examination with 'od -v'. >> >>> In an attempt to kick off the reshape operation, I issued: >>> >>> # setenforce 0 >>> >>> to turn off SELinux enforcement. That didn't help - the reshape sat >>> still, showing no progress. >>> >>> Then I issued: >>> >>> # mdadm --stop /dev/md6 >>> >>> which of course interrupted the reshape operation. It also threw up a >>> bunch of error messages, which you can find in the dmesg.txt file >>> found at https://app.box.com/s/3pksam3c7n79anpnzvsrwekzqwtsvlf6 -- >>> look for the words "cut here". It looks like a segfault or other >>> runtime error: >>> >>> [ 796.84193] WARNING: CPU: 0 PID: 1444 at mm/backing-dev.c:372 >>> bdi_unregister+0x38/0x50() >>> >>> I then tried to restart the grow operation, without SELinux' help, and >>> got the error message in the subject. >>> >>> First, I goofed, and tried the assemble without the backup file: >>> >>> # mdadm --assemble /dev/md6 /dev/sdc4 /dev/sdd4 /dev/sdg4 /dev/sdh1 /dev/sdj1 >>> >>> [ 1966.030411] md: md6 stopped. >>> >>> mdadm: Failed to restore critical section for reshape, sorry. >>> >>> # mdadm --assemble /dev/md6 /dev/sdc4 /dev/sdd4 /dev/sdg4 /dev/sdh1 >>> /dev/sdj1 --backup-file=/home/gwr/c/grow_md6.bak >>> >>> [ 2242.492370] md: md6 stopped. >>> >>> mdadm: Failed to restore critical section for reshape, sorry. >>> >>> # mdadm --assemble /dev/md6 /dev/sdc4 /dev/sdd4 /dev/sdg4 /dev/sdh1 >>> /dev/sdj1 /dev/sdi1 --backup-file=/home/gwr/c/grow_md6.bak >>> >>> [ 2403.741995] md: md6 stopped. >>> >>> mdadm: Failed to restore critical section for reshape, sorry. >>> >>> I ran an mdadm --examine on all my RAID partitions; the file is at >>> https://app.box.com/s/9x2n2wc42i1wqzd1ayrt8ta6cyldrr6i. Of note in >>> that file: the "Reshape pos'n" is 0 on all six drives. I take that to >>> mean that the reshape operation never really got started. >>> >>> Is my next step to add the --invalid-backup switch? If not, what >>> recommendations might you have to fix this? >> >> I tried the --invalid-backup switch this morning without success: >> >> md: md6 stopped. >> md: bind<sdc4> >> md: bind<sdg4> >> md: bind<sdh1> >> md: bind<sdj1> >> md: bind<sdi1> >> md: bind<sdd4> >> md/raid:md6: reshape_position too early for auto-recovery - aborting. >> md: pers->run() failed ... >> mdadm: failed to RUN_ARRAY /dev/md6: Invalid argument >> md: md6 stopped. First question - is anyone besides me seeing these messages? I see my postings in the archive at http://marc.info/?l=linux-raid&r=1&b=201508&w=2 so I assume they're going out to the distro list, but they have not been sent back to me nor have I received a reply. I had a rough time posting the initial message last night due to Gmail's default use of HTML, which the linux-raid mailer-daemon does not like -- it only accepted the message when I turned Gmail's "Plain text mode" on. Second, I see in http://lwn.net/Articles/565591/ that mdadm 3.3 has an undocumented option: "--assemble --update=revert-reshape" can be used to undo a reshape that has just been started but isn't really wanted Since I'm on mdadm v3.3.2, I assume I still have that option. Is it worth trying in my situation? Thanks for any advice. George -- 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
