On Dec 5, 2006, at 12:13 PM, Rutger Bevaart wrote:
There were 2 targets created of 4GB, online and used. They were
connected to by static configuration using a local network
interface (nge1), and mounted as a local filesystem.
I created one additional target of 120GB which never came online
(at least not within a few hours). When I attempted to delete it,
in time this is roughly when it crashed. I cannot be sure it was
that what triggered it...
Given the stack trace, it was the delete that caused the core. Some
quick background on how the target creates logical units. The
administrator will request that the daemon create a LU of some size,
in your case that was 120GB. The daemon does a quick check to see if
there's room of the device for a LU of that size. If so, it returns
success and starts a separate thread to initialize the LU. This
thread basically has to write every block in the file to guarantee
the space is available. If it doesn't perform this operation the file
would be hole-y, this is called thin-provisioning which has pros and
cons.
Now, as you've seen this can take quite a long time to initialize the
LU. When you deleted the LU, the daemon needed to shutdown that
thread and does so by sending the thread a message to exit. The
thread will send an ack message so that the main thread knows that it
got through and then exit. The main thread then does some cleanup.
This is where there's some problem which has escaped our test suites.
@Rick,
Given some time I can do some ZFS work, but if I understand
correctly I cannot use a partition as ZFS right? I might be able to
break the mirror and use the second disk for ZFS - if that helps
debugging...
I use ZFS on several of my machines which only have a single disk. I
use the partition which is normally setup by the Solaris install
program as /export/home. I remove /export/home from /etc/vfstab,
unmount it, and then run:
zpool create -f playground c0t0d0s7
There's no redundancy, but for my testing that's fine.
Since I now know where in the code the problem is happening and what
you where doing when this occurred, I should be able to reproduce the
problem.
My suggestion to use ZFS was more from the standpoint that UFS will
not give you the performance that you might seek. I look at UFS/SVM
in the same way as my old 1970 Dodge D100 pickup truck. It was a good
old truck that I loved, but it would haul to much nor get anywhere
real fast. ZFS is like my new 2006 Dodge 2500 with the Cummins Diesel
engine. This truck pulls my 8000 pound trailer like it's not even there.
If you have a reason to run UFS/SVM then do so.
This message posted from opensolaris.org
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
----
Rick McNeal
"If ignorance is bliss, this lesson would appear to be a deliberate
attempt on your part to deprive me of happiness, the pursuit of which
is my unalienable right according to the Declaration of
Independence. I therefore assert my patriotic prerogative not to
know this material. I'll be out on the playground." -- Calvin
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
