On 2021-02-15 22:31, Michael D. O'Brien wrote:
Hi, I have a single mdadm raid6 in a 56-drive raid60 (7x8) with a
kernel thread stuck at 100% cpu. The stuck thread typically happens
during array checks, but is not the resync thread - md122_raid6 is at
100% cpu, whereas md122_resync is at ~0%. When this happens, the
reported sync speed drops until it reaches 4K/sec. Setting sync_action
to idle gets stuck.
iostat shows backing devices aren't doing anything i/o wise, SMART is
clean for all member drives, and dmesg doesn't say anything useful
(until the thread is hung for a long time, then it tells me as much -
I'll post that message when the current issue times out). A reboot
typically clears the issue, but takes quite a long time, as the raid
60 is the backing device for a bcache device (with an optane cache)
that has a large mounted xfs file system in place.
I figured I could strace the process, but I learned that's impossible
with kernel threads :)
[...]
Hello Michael,
This sounds pretty much the same what we have experienced whilst
checking raid6 assemblies.
The issue is actively tackled in the moment, c.f the "[PATCH V2] md:
don't unregister sync_thread with reconfig_mutex held" thread.
And more details in the link:
https://lore.kernel.org/linux-raid/5ed54ffc-ce82-bf66-4eff-390cb23bc1ac@xxxxxxxxxxxxx/T/#t
Kind regards,
Thomas
--
Thomas Kreitler - Information Retrieval
kreitler@xxxxxxxxxxxxx
49/30/8413 1702
