Re: [HPDD-discuss] Strange Issue, No space left, where there actually over 500TB
Hi Rick,
Yes I am on same page with you on restriping. Also I found out that we have about 46 files
that are sized over 1 or 2 TB that caused the OST's to fill up sooner than others.
This probably explains why handful of 22TB OST was soon filled up than others.
Robinhood usage profile helped me dig into this quicker!!
Thank you for your help.
Amit
> On Jun 1, 2015, at 1:30 PM, Kumar, Amit <ahkumar(a)mail.smu.edu> wrote:
>
> Correction: If a file already exists then I can't restripe it unless I
redistribute
it. So I guess I cannot salvage any running jobs. I hope I got this right?
Just to clarify, when I spoke of “redistributing data”, I was just talking about
restriping the file. If a file is actively being used, you probably should not try
to restripe it (unless you are positive that the file is only being read and not
written). To restripe a file, you just create a new file with a different stripe
count and then copy the data:
lfs setstripe -c $new_stripe_count NEW_FILE cp OLD_FILE NEW_FILE mv
NEW_FILE OLD_FILE
If the file is actively being written, then you run the risk that some data won’t
be copied.
--
Rick Mohr
Senior HPC System Administrator
National Institute for Computational Sciences http://www.nics.tennessee.edu