lord@xxxxxxx wrote:
>> i just tried the current state of the sgi xfs cvs tree, got it
>> compiled and running fine but after mounting a filesystem and
>> copying lots of stuff to it i got lots of
>>
>> pagebuf_kiobuf_io got error -38
>>
>> errors and the machine was unuseable a bit later - is this a
>> known problem at the moment ?
> I suspect you are using ide drives and turned on the use kiobuf
> option. Try rebuilding with the use kiobuf config option off and
> try again.
thanks a lot - that it was ... i did some more testing of it over
the weekend and it looked much more promising than last time i tried
(~ 8 weeks ago) ... got it to hang the machine only once (sorry no
debug output available) - but sofar it worked quite well (even nfs
on top of it seems to run quite good due to my limited testing) -
thanks for all the good work so far ...
i today started to use it here at work too - built the kernel, mkfs'ed
the fs and mounted it - copied the / filesystem over to it - and while
doing this i got:
kernel: I/O Error Detected. Shutting down filesystem: ide0(3,5)
kernel: Please umount the filesystem, and rectify the problem(s)
kernel: PCD: pagebuf_bmap error -5 pb_flags 0x10010002
but the disk did not have any hw error (no ide error to see and the
disk is fresh and fine) - retried it (reboot, mkfs, mount, cp) and
it works now ... will run it in "semi-production" here now - maybe
i can help finding some bugs :-)
also some questions:
* are there any plans to move it towrds a newer test version of the
2.4.0 kernel ? (because i noticed a lot of the test1 vm problems
in my tests and for using it on a really used machine it would
be better to have a kernel without all those problems - which are
i think partly fixed in the later versions)
* is there any timeframe for a (alpha or beta) working
xfsdump/xfsrestore ?
also somthing i noticed - looks like one should add the "read-write"
option to the lilo entry if the root is xfs - otherwise it has
problems doe to trying to replay the log on a ro filesystem (paniced
in this case once at home) - this might be (if it is correct)
interesting for other who wanted to try this
t
--
thomas.graichen@xxxxxxxxxxxxx
clustering & security innominate AG
networking people
fon: +49.30.308806-13 fax: -77 web: http://innominate.de pgp: /pgp/tgr
|