Closed agarwal closed 10 years ago
From: Karl Ward kw1213@nyu.edu Date: Fri, Jul 26, 2013 at 12:46 PM Subject: Backup policy for /data/cgsb To: Lior Atar la466@nyu.edu, Sreedhar Manchu hpc@nyu.edu, Ashish Agarwal ashish.agarwal@nyu.edu
Back when we first started backing up /data/cgsb, we decided to only backup certain directories in /data/cgsb. Perhaps that decision should be reconsidered now. Ashish, take it away.
Karl Ward
From: Lior Atar la466@nyu.edu Date: Fri, Jul 26, 2013 at 2:10 PM Subject: Re: Backup policy for /data/cgsb To: Karl Ward kw1213@nyu.edu Cc: Sreedhar Manchu hpc@nyu.edu, Ashish Agarwal ashish.agarwal@nyu.edu
Hello all,
Due to the nature of TSM (tivoli) configuration, we need explicitly list each directory to include. This is done for two reasons,
-lior
From: Ashish Agarwal ashish.agarwal@nyu.edu Date: Fri, Jul 26, 2013 at 2:34 PM Subject: Re: Backup policy for /data/cgsb To: Lior Atar la466@nyu.edu Cc: Karl Ward kw1213@nyu.edu, Sreedhar Manchu hpc@nyu.edu
On Fri, Jul 26, 2013 at 2:10 PM, Lior Atar la466@nyu.edu wrote:
- Otherwise we would have to then exclude every other directory on lustre ( major pain)
I don't get this. Right now you can say "backup /scratch/cgsb/aa144, /scratch/cgsb/bonneau, etc." It should be reducible and simpler to say "backup /scratch/cgsb". In both cases you are specifying one or more sub-directories of /scratch to include.
- That way each directory under /data/cgsb then looks like another FS to tsm and then allows multi-threaded scans of those directories ( resulting in faster backups and queries on TSM end)
This will be resolved once we tar up the gencore raw directories, but that's held up until I get an answer to my previous email below.
-Ashish
From: Ashish Agarwal ashish.agarwal@nyu.edu Date: Mon, Aug 26, 2013 at 6:05 PM Subject: Re: Backup policy for /data/cgsb To: Lior Atar la466@nyu.edu Cc: Karl Ward kw1213@nyu.edu, Sreedhar Manchu hpc@nyu.edu, Sebastien Mondet sebastien.mondet@nyu.edu
Hi. I'd like to revisit the request below now that the file count has been drastically reduced. Can you please institute a full backup of all of /data/cgsb. This would be much better and give us confidence we aren't accidentally omitting directories as they get added.
As of now, we have to tell the HPC team each sub-directory under
bowery:/data/cgsb
that we would like to have backed up. Thus, if we add a new sub-directory, we have to remember to tell the HPC team to add that to the backup script.This is unsatisfactory because: