[Bioclusters] resources on administering clusters

Joe Landman bioclusters@bioinformatics.org
02 Apr 2002 11:56:36 -0500


Confusion over my language as far as I can tell, and one mistake on my
part.  If you want to parallelize your swap, all your priorities need to
be identical. 

Setting the swap as listed in the document fill up one bit of the swap
before the others.  What was incorrect was my ordering.  Roger is
correct about pri=32767 being used before pri=0.  

See 

	man 2 swapon 

for the specifics.

So change the pri=1 to pri=3 for a 4 way "stripe" (really a round robin
method and not a true stripe).



On Tue, 2002-04-02 at 11:14, Roger Pettett wrote:
> On 26 Mar 2002, Joe Landman wrote:
> 
> > On Tue, 2002-03-26 at 10:15, Andrew Shewmaker wrote:
> > > Jeff Layton wrote:
> > > 
> > 
> > > Set multiple swap partitions to the same priority
> > > 
> > > |
> > > /dev/sda2	none	swap	sw,pri=3	0	0
> > > /dev/sdb2	none	swap	sw,pri=3	0	0
> > > /dev/sdc2	none	swap	sw,pri=3	0	0
> > > /dev/sdd2	none	swap	sw,pri=1	0	0
> > 
> > Almost...  Set all the pri=0.  Otherwise swap is not parallelized.  If
> > one partition is set to pri=1, and the rest set to pri=3, swap will fill
> > the pri=1 up before attempting to use the pri=3 resources.    
> > 
> 
> This isn't correct - pri can be any value between 0 and 32767 with 32767 
> being the highest priority. Swap is parallelised between any partitions 
> with the same 'pri' value, ordered from highest to lowest.
> 
> R.
> 
> -- 
> +---------------------------------------------------------------+
>  Roger Michael Pettett           Email: rmp@sanger.ac.uk
>  Project Leader (Web Systems),   Web: http://www.sanger.ac.uk/
>  The Sanger Institute                 http://www.yourgenome.org/
>  Wellcome Trust Genome Campus         http://www.ensembl.org/
>  Hinxton                       
>  Cambridge CB10 1SA
> +---------------------------------------------------------------+
> 
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> http://bioinformatics.org/mailman/listinfo/bioclusters
-- 

Joseph Landman, Ph.D.
Senior Scientist,
MSC Software High Performance Computing
email		: joe.landman@mscsoftware.com
messaging	: page_joe@mschpc.dtw.macsch.com
Main office	: +1 248 208 3312
Cell phone	: +1 734 612 4615
Fax		: +1 714 784 3774