Hi Ti,
Can you or did you try restarting the servers (after shutting down
the clients)?
-sam
On Aug 26, 2006, at 1:04 PM, Ti Leggett wrote:
Right. Here's the deal, when I started this whole process each one
of those nodes that is now full had like 28G free. So by copying
the file to a another location, removing the old file, and then
moving the file back in place did not result in a net increase of 0
like I think it should. I know which files are the largest and I
could probably move those off of the PVFS2 volume to local disk
temporarily, but if by trying to rebalance these files I'll end up
losing space, then I'm doing something wrong or something is going
wrong in PVFS2. Either way, I need to find out why this process is
using space instead of resulting of a net usage of 0 like it should.
On Aug 26, 2006, at Sat,Aug 26, 12:06 PM, Robert Latham wrote:
On Sat, Aug 26, 2006 at 11:14:45AM -0500, Ti Leggett wrote:
Well I was hoping I was doing something wrong :)
Here's the statfs output.
server: tcp://nightcrawler.ci.uchicago.edu:3334
handles available: 858976089
handles total : 858993459
bytes available : 5894144
bytes total : 246131359744
mode: serving only I/O data
This is not so good: only 6 megs of free disk space.
server: tcp://rogue.ci.uchicago.edu:3334
handles available: 858976089
handles total : 858993459
bytes available : 11202560
bytes total : 246131359744
mode: serving only I/O data
11 megs here isn't so hot either
server: tcp://wolverine.ci.uchicago.edu:3334
handles available: 858976089
handles total : 858993459
bytes available : 9928704
bytes total : 246131359744
mode: serving only I/O data
space is a little tight here, too
So we need to focus on those three first. nightcrawler is the reason
you only have 5*8=40 MB free, but the other two aren't helping much
either: Are those three the original servers? Do they have a log
file or anything else consuming space in the pvfs2 storage-space
that's not part of PVFS2?
Do we have a good way to reverse-map data files on disk to pvfs2
files? If so, you could find what the largest datafiles on
nightcrawler, rouge, and wolverine were, and then free up some space
by moving them somehwere temporarily.
I think we can figure this out. Just might take a few rounds.
==rob
--
Rob Latham
Mathematics and Computer Science Division A215 0178 EA2D B059 8CDF
Argonne National Labs, IL USA B29D F333 664A 4280 315B
_______________________________________________
Pvfs2-developers mailing list
Pvfs2-developers@beowulf-underground.org
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
_______________________________________________
Pvfs2-developers mailing list
Pvfs2-developers@beowulf-underground.org
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers