On Mon, 7 Nov 2005, Charles Loomis wrote:
> Hello,
>
> Our SE at LAL (grid11.lal.in2p3.fr) is full because of recent activity
> by the atlas and biomed VOs. However, looking at the utilization of the
> space by VO:
>
> dteam: 39 GB
> biomed: 14 GB
> atlas: 5 GB
>
> dteam dominates the usage. Having dteam using 2/3 of the disk is not
> reasonable. These files are around 5600 small files which have LFNs like:
>
> lfn:dteam07-19696-dstorm-S11003-I3.data
>
> or
>
> lfn:dteam07-mdstorm-S10063-I0099-CEgrid10-WNgrid02-oF49-32293-041006-204242.mds
Hmmm, those names suggest that a certain well-known test suite was used... :-)
> most of which are from last October and November. They take a large
> amount of space because of the block size.
>
> I can of course transfer these files (and the problem) to another SE on
> the grid. However judging from the names and the dates, I'd guess that
> these are not useful any more and would prefer to delete them.
>
> Is there anyone who believes that these files are still useful?
Just delete them.
> After the clean, I will up enable a quota for dteam. What is a
> reasonable size to support the testing activities of dteam? 5GB, 10GB?
I would say that dteam shall leave no files behind at all, but jobs may
fail to clean up after themselves, so at least a few GB should be allowed,
and a cron job might throw away old and/or big files.
But then again, these days a 100 GB disk costs less than 100 euros,
so you might rather want to enlarge your SE a bit...
|