>Hi all,
>
>as of 21:57 Dec. 16, /home/blast has 550k free space on it. the last time
>we discussed the same issue was Dec. 1st.
>
>I dare to suggest that the use of /home/blast be discussed in one of the
>so many meetings and a solution be worked out that would prevent the same
>situation to reemerge in at least the next six months.
>
>I can see a few approaches,
>
>1. a quota system that limits the space of each user. since Ernie has made
>/scratch1/bud18 fully backed up, I do not see any reason to store big
>amount of data on /home/blast.
Before implementing a quota system (if that is decided), we must get the heavy
users to reduce their usage to some reasonable amount.
My system to automatically keep /scratch1/bud18 backed up has been in place
since Dec 6th. As of today, not even 1 file has been moved from /home/blast
to this area. I agree that a quick/simple solution for freeing space is to
move large data files to this area. /scratch1/bud18 should be available to
all Linux systems utilized by Blast users.
>2. in crease the space in /home/blast. a 40G hard drive costs less than
>$100 retail. add a $100 to fully back up. People will still need to use
>restraint though as a Monte Carlo session could easily eat up gig bytes of
>data.
The cost would be quite a bit more than that The /home areas on the server
are configured as Raid 5, which means a minimun of 3 disks are needed. Plus
they are SCSI disks which are conciderably more money. However, I believe I
may have some disks (18 gig each) that were used in another Raid server no
longer in use. 3 of them in a Raid 5 configuration essentially gives you
32gig of space.
Another solution to increase space would be to repartition the existing Raid
set disks that /home/blast occupies. There is 3 gig I from /home/sample (or
I could archive /home/sample and take all 5 gig of it).
All alternatives involve a good amount of down time of /home/blast (and some
of the server itself). It's tough to give a good estimate of how long though.
Also, as Doug mentioned in one of his replies, the amount of time needed to do
backups is an issue (only for FULL backups). Currently it takes approximately
24 hours to complete a FULL backup. I don't believe any of the solutions above
would increase this unreasonably. Anything that is moved to /scratch1/bud18
is not part of this issue because it is a separate backup server than the one
I use for gerneral Unix backups.
>3. deprecate some inactive users. However, the decision about who to
>remove will need to be made by the administration level.
Once this is decided I can just move all their files to /scratch1/bud18.
>4. not quite realistically, maybe a different net work topology.
I'm not sure what you mean by this. The network topology has nothing to do
with /home/blast space issues.
>BTW, the blasttalk archive is being swamped by messaged from crond daemon
>seemingly originated from blast02
Not sure what you mean by this exactly. I'll check it out.
Ernie
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:32 EST