[PLUG] killing "D" state processes?
Russell Senior
seniorr at aracnet.com
Tue May 14 23:19:20 UTC 2002
>>>>> "Shannon" == Shannon C Dealy <dealy at deatech.com> writes:
Shannon> On 14 May 2002, Russell Senior wrote:
Russell> I've been having a little trouble with a backup process.
Russell> This morning an 'mt' hung on me in a "D N" state, which
Russell> according to "man ps" decodes as Wing Attack Plan R, R for
Russell> Robert ... Oops! No, sorry. Those weren't today's codes.
Russell> It decodes as D = "uninterruptible sleep (usually IO)" and N
Russell> = "low-priority task". A "kill -9 <pid>" doesn't work. The
Russell> tape wasn't moving, I ejected the tape and reinserted it, but
Russell> mt is still stuck. Do I have to reboot the sucker? Or is
Russell> there some incantation like: "Peace on Earth" or "Purity of
Russell> Essense" to fix it?
Shannon> I used to get these years ago during tape backups (including
Shannon> the failure of the kill -9), it turned out to be a kernel
Shannon> bug/race condition that would happen on rare occasions, [...]
Shannon> What kernel version are you running?
"uname" says 2.4.16. Oddly, about 5 minutes after I sent the message,
the process went away and I could at least "mt status" the drive.
But, when I tried to rewind, it hung again. I am getting messages in
the logs that indicate some kind of failure (either communication or
hardware, not sure yet). They look like:
May 14 14:57:41 coulee kernel: scsi : aborting command due to timeout : pid 13945266, scsi0, channel 0, id 5, lun 0 Rezero Unit 00 00 00 00 00
May 14 14:57:41 coulee kernel: scsi0: Aborting CCB #7064897 to Target 5
May 14 14:57:41 coulee kernel: scsi0: CCB #7064897 to Target 5 Aborted
May 14 14:57:41 coulee kernel: st0: Error 26030000 (sugg. bt 0x20, driver bt 0x6, host bt 0x3).
I am thinking a nice gentle reboot at this point might clear it up.
--
Russell Senior ``The two chiefs turned to each other.
seniorr at aracnet.com Bellison uncorked a flood of horrible
profanity, which, translated meant, `This is
extremely unusual.' ''
More information about the PLUG
mailing list