[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] tap:aio not working...



Marcel Ritter schrieb:
Hi again,


...
If also found this behaviour when using sparse qcow files:

qcow-create -r 20480 /proj.stand/image.qcow
xm block-attach 0 tap:qcow:/proj.stand/image.qcow xvda w 0
... <insert test above here> ...

works fine, but without "-r (reserve)" flag, it fails miserably:

qcow-create 20480 /proj.stand/image.qcow
xm block-attach 0 tap:qcow:/proj.stand/image.qcow xvda w 0
... <insert test above here> ...
.... < this takes hours ... or more? Never finished for me)

Update on that (qcow, without reserve):
I was too impatient on that, after waiting quite some time I got this:

while true; do time dd if=/dev/zero of=dummy bs=1M count=10240; done
10240+0 records in
10240+0 records out
10737418240 bytes (11 GB) copied, 3851,42 seconds, 2,8 MB/s

real    64m11.684s         <- took over 1 hour!!!
user    0m0.032s
sys     0m12.569s
10240+0 records in
10240+0 records out
10737418240 bytes (11 GB) copied, 3928,04 seconds, 2,7 MB/s

real    69m0.305s
user    0m0.040s
sys     0m14.485s
10240+0 records in
10240+0 records out
10737418240 bytes (11 GB) copied, 336,601 seconds, 31,9 MB/s

real    5m44.421s
user    0m0.012s
sys     0m14.021s
10240+0 records in
10240+0 records out
10737418240 bytes (11 GB) copied, 111,804 seconds, 96,0 MB/s

real    1m54.282s          <- now, less than 2 minutes
user    0m0.032s
sys     0m13.189s
10240+0 records in
10240+0 records out

Looks like expanding the qcow file the first time takes *very* long, after
that it speeds up considerably (2,8 MB/s -> 96 MB/s), so my next try will be
to run that test again on tap:aio: with sparse file ....

Currently running the same test for tap:aio: again ... running for more than 3 hours now ...
So this is (in short) what works (and what not):

type:      sparse:      works:
file:/             no               yes
file:/             yes             yes
tap:aio:/      no               yes
tap:aio:/      yes             no
tap:qcow:/   no              yes
So this should read:
- tap:qcow:/   yes            no
+tap:qcow:/   yes            yes
So for me it looks like blktap (or some other component) gets into trouble when
sparse files are used...

Any idea what's the cause of that?

BTW: The system I'm currently testing on is SLES 10 SP2.
Bye,
  Marcel

--
----
Dipl.-Inf. Marcel Ritter
Linux/Novell
Regionales Rechenzentrum Erlangen
Tel: 09131 / 85-27808
E-Mail: Marcel.Ritter@xxxxxxxxxxxxxxxxxxxx
----
Unix _IS_ user friendly... It's just selective about who its friends are.



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.