[Halla_running] Triton /work migration

[Ole Hansen]

Hello everybody,

FYI, the Computer Center will migrate the /work/halla/triton disk area
from the Lustre disk to a new file server on *Monday, May 7* between
approximately 8-10 am.  The new server promises more space and speed.

This is a low-risk operation. Most of the data have already been
transferred. What is left to do is a final sync and the switchover of
the mount points. The old disk area (on Lustre) will continue to be
available for a few days as a backup, in the unlikely case that anyone
notices anything missing.

What this means for you:  Please plan not to use /work/halla/triton next
Monday morning.  Specifically:

- Close all open files on that disk before the migration. Quit editors
that have opened source files and scripts.

- On Monday morning, do not run farm jobs or run replay jobs (on adaq or
elsewhere) that read from or write to /work/halla/triton. If you have
long-running jobs, please stop submitting them early enough so that they
can finish before Monday.  Anything that is written to
/work/halla/triton during the migration is at risk of not being
transferred properly.  The triton areas on /cache and /volatile are not
affected by this move, so feel free to keep using them.

- If you have source files, scripts, databases etc. on
/work/halla/triton that are particularly valuable to you, now would be a
good time to make a personal backup.  (You were going to do that anyway,
weren't you?)  You can keep small backups on /home and write all
backups, small and large, to /mss/home/<your CUE user name>/<some
subdirectory if applicable>/ using jput.

- After the migration, verify that all your files are there and intact.
Here's how to record and check the integrity of all files in a directory
tree:

(1) Calculate and save checksums:

% cd /work/halla/triton/mydir/mysubdir
% find . -type f -exec md5sum {} \; > ~/mysubdir.md5

This may run for a while. When done, ~/mysubdir.md5 should contain one
checksum and file name per line. Any changes that happen from now on
will show up when you run the verification:

(2) Verify checksums:

% cd /work/halla/triton/mydir/mysubdir
% md5sum -c ~/mysubdir.md5 | grep -v ": OK"

If you get any output, don't panic. (a) the originals are still there,
under /work/halla/triton.old/, and (b) some files may have legitimately
changed, e.g. because you recompiled something etc. Nevertheless,
investigate.

You will be able to tell that the disk migration has been completed by doing

% df /work/halla/triton

If you see "scifs17exp-eth:/expphyvol/halla/triton" or similar under
"Filesystem", then you are using the new server. If you see something
containing "lustre", you are still on the old server.

Please let me know if you have any questions or run into problems.

Ole