>Hi,
>
>instead of yanking users out of /home/blast, I think it might be more
>helpful to create an area in the scratch discs that is backed up. don't
>need to be all that often, maybe once a month or even once a quater.
>
>I remember it happened to both Adrian and Aaron when considerable portion
>of their work were lost due to scratch disc failure. If the scratches on
>bud10 fail right now, I estimate about a month set back on my work. All
>the ntuples must be refiltered, all the monte carlos must be regenerated,
>plots remade, latest changes to codes must be recalled and reprogrammed
>from the last point saved in CVS.
>
>I believe this is partly the reason why people are reluctant to use the
>scratches but instead put significant amount in home/blast.
I do have a tape drive connected to bud18 that was used for archiving the
BLAST data and /scratch1/bud18 is virtually empty. I'll setup a cron job to
do a full backup of /scratch1/bud18 once a week. To keep it somewhat automated
I'll write over the same tape each time, but will archive the tape every so
often ( once a month or 2 ). Plus, I'll do daily backups to disk files and
store them on /scratch/bud18. /scratch/bud18 is a different physical disk than
/scratch1/bud18, so if /scratch1/bud18 became corrupt we could still fully
recover. I'll post to the list when the system is functional and hopefully
users will move large data files to the area.
>also with time, a lot of temporary files could be left on one's home dir.
>they do not show with "ls" command, for example:
> .autosave-*
> .save*
> .backup-*
> .nfs*
>they should be cleaned up regularly.
As well as core files. I do periodically scan for core and .nfs* files and
remove them. I was unaware of .autosave-*, .save*, and .backup-* files though.
Ernie
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:32 EST