Solution to BLCR Segmentation Bug

Blog post edited by Anonymous - "Migrated to Confluence 4.0"

After a few months since our install of RHEL6 and my original post reporting the problem with BLCR:

• https://wiki.hpcc.msu.edu/x/c6LT

We have found and tested a solution that works on our system. Here is a link to a detailed description of the Prelink problem/BLCR and how to fix it:

• http://en.wikipedia.org/wiki/Prelink
• https://upc-bugs.lbl.gov/blcr/doc/html/FAQ.html#prelink

Note, for a while I was not convinced that Prelinking was a problem or the only problem because my BLCR script used the --save-all option. Using this option seems to have also messed up our system. Removing the --save-all option is not a problem on our system since all of our compute nodes have same image and directory structure.

This has already made a lot of users on our HPC system happy because it allows them to get around our walltime limits and makes their jobs more robust to node failure.

• Dirk

View Online

Blogpost migrated from ICER Wiki using custom python script. Comment on errors below.