[Lustre-discuss] Lustre::LFS + Lustre::Info (inc. lustre-info.pl) available on the CPAN

Tina Friedrich Tina.Friedrich at diamond.ac.uk
Thu Jul 29 16:55:21 UTC 2010


Thanks Andreas,

that explains that.

So the warning can be made to go by changing line 183 in lustre-info.pl from

printf(", %s=%5.1f R/s",$type,$stats->{$type}/$slice)

to be

if (exists $stats->{$type})
	{ printf(", %s=%5.1f R/s",$type,$stats->{$type}/$slice); } 


(patch attached)

Tina

Andreas Dilger wrote:
> On 2010-07-29, at 09:08, Frederik Ferner wrote:
>> Note the 'Use of uninitialized value in division...' errors. Looking at 
>> the code it seems the value for 'setattr' is missing from the stats file 
>> for some of our OSTs. Looking at the stats file, indeed the setattr line 
>> is missing for some OSTs.
>>
>> Has anyone seen this before? What could have caused this?
> 
> The statistics code for OBD devices is generic.  For operations that have never been done on a particular target there is no stats value printed.  Otherwise, the "stats" file would be 60 lines long and mostly be filled with counters that are all "0".
> 
> Cheers, Andreas
> --
> Andreas Dilger
> Lustre Technical Lead
> Oracle Corporation Canada Inc.
> 
> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss at lists.lustre.org
> http://lists.lustre.org/mailman/listinfo/lustre-discuss
> 


-- 
Tina Friedrich, Computer Systems Administrator, Diamond Light Source Ltd
Diamond House, Harwell Science and Innovation Campus - 01235 77 8442
-------------- next part --------------
--- lustre-info.pl	2010-07-29 17:45:10.000000000 +0100
+++ lustre-info.pl_new	2010-07-29 17:45:54.000000000 +0100
@@ -180,7 +180,8 @@ sub loop_ost_stats {
 			
 			# Add some 'metadata' info
 			foreach my $type (qw(create destroy setattr preprw)) {
-				printf(", %s=%5.1f R/s",$type,$stats->{$type}/$slice);
+				if (exists $stats->{$type})
+					{ printf(", %s=%5.1f R/s",$type,$stats->{$type}/$slice); }
 			}
 			print "\n";
 		}


More information about the lustre-discuss mailing list