Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Wed, Aug 22, 2007 at 03:07:13PM +1000, Maciek Olczak wrote:
> I know that sometimes requests may be queued almost
> indefinetely, especially on MD, which strips off the
> BIO_RW_SYNC bit from requests passed down to it, I
> wrote a patch for that myself.
>
>
>
> Am I correct thinking this is not a part of rel. 0.7.24 ?
nothing to do with drbd, it is a kernel patch,
and it has been included in upstream long before 2.6.22.
(it is even in 2.6.16.something,
though aparently it never made it into 2.6.18.y...)
> maybe it helps to say "sync", or enable sysrequest
> ("echo 1 > /proc/sys/kernel/sysrq"), and then trigger
> an emergency sync ("echo s > /proc/sysrq-trigger").
>
> Thanks for this hint. I'll put it in and see if clears these "limbo" requests.
>
> BTW, the other server's hang (see above) was for the same reason. On both
> servers, every 3 minutes I dump /proc/drbd into a log. From there I can confirm
> that on each occasion there were outstanding meta dev requests before
> disconnect was initiated. Here is what happened in this event:
2.6.22 has changed io request handling again to reduce stack usage of
stacked block devices. we may have potential for a deadlock in the
generic_make_request code, I'll have to look deeper in that.
so for now, maybe rather stay away from 2.6.22 until we verified
"compatibility" of drbd.
--
: Lars Ellenberg Tel +43-1-8178292-0 :
: LINBIT Information Technologies GmbH Fax +43-1-8178292-82 :
: Vivenotgasse 48, A-1120 Vienna/Europe http://www.linbit.com :
__
please use the "List-Reply" function of your email client.