SiteReportMay19

lsm-get failures

• lsm-put issues continue to dominate failure rates
• Charles has been doing diagnostics based on logfile data
• #movers per pool: 500.
• lsm supposed to handle its own timeouts
• Pilot time-out to be increased from 30 minutes to 90 minutes

Kernel update testing

• This is to address the soft-locking kernel error we've seen.
• Built latest kernel 2.6.33-4 for one of MWT2_UC's worker nodes
• uct2-c201-209 ready to install - will be running jobs
• Considered also SL 5.5 kernel via rpm (2.6.18-xx)
• Charles studied optimizations during build - lots of interesting findings, to feedback to SL.

Squid

• Squid now updated at IU and UC

ANALY_MWT2_X

• Testing local site mover prototype from Charles using this analysis queue
• Checksum for lsm-puts handled by reading back from worker node; no dependence on Bestman

Bringing more storage online

• Using 707TB of 915TB (77%)
• Current space report: http://www.mwt2.org/sys/space
• Expect to bring another server (with 6 MD1000 shelves, 2 TB drives, 180 TB raw) within the week

Space token graphing in Cacti

• graph_image.php.png:
  

Network and jobs activity (UC)

• graph_image-1.php.png:
  

• graph_image-2.php.png: