test: fixup and improvements for rbd-mirror test

[trociny]

Signed-off-by: Mykola Golub mgolub@mirantis.com

[trociny]

To troubleshoot a test failure I use the following commands:

From src dir start the test in NOCLEANUP mode (so it possible to check its state after failure) and with TEMPDIR pointing to known location):

RBD_MIRROR_NOCLEANUP=1 RBD_MIRROR_TEMDIR=/tmp/tmp.rbd_mirror ../qa/workunits/rbd/rbd_mirror.sh

After failure cd to TEMPDIR and check status:

cd /tmp/tmp.rbd_mirror
less rbd-mirror.log
rbd --cluster remote -p mirror journal status --image test
ceph --admin-daemon rbd-mirror.asok help
...

Run cleanup:

cd - #SRCDIR
RBD_MIRROR_TEMDIR=/tmp/tmp.rbd_mirror ../qa/workunits/rbd/rbd_mirror.sh cleanup

[jdurgin]

Could you add that comment on how to use it to the script (perhaps a comment on using the script at the top of the file)?

[jdurgin]

looks like this isn't used yet

[trociny]

Yes, I was going to use this to test live image replayer removal after an image is removed (journaling is disabled). But right now I can't remove the image when it is replayed due to "Device or resource busy" error. Still I think it is desirable feature and we will be able to do this in some future, and this function will be useful then.

[jdurgin]

makes sense to leave it in then. I think Jason fixed that EBUSY error with his aio_notify patch that's now in master.

[trociny]

Rebased after adding "how to use" into the scrip.