perfSONAR Announcements

[Jason Zurawski <>]

Greetings;

We have received reports recently of users seeing kernel panics when 
installing new software, or rebooting after a long time of being in use.  In 
triaging and debugging this, the root cause seems to relate to a lack of 
space in the /boot partition on the node.  CentOS changed the way that the 
initial partitioning was handled in version 6 of the operating system, 
meaning that our project has less ability to control the space allotment 
outcome via the installation process, and the result is that the limited 
space in /boot can only support a small number of kernels.  

We have written some instructions on how to mitigate this on the FAQ site:

http://psps.perfsonar.net/toolkit/FAQs.html#Q72

For now we are recommending that admins check the space available in /boot, 
and use yum to remove older kernels that are no longer needed (saving one 
backup is always a good idea).  Future release announcements will also 
feature a reminder of this, until the project can figure out a better 
solution to either alarming on this condition, or altering the disk layout so 
it is no longer a serious concern.  

Thanks for your support of perfSONAR, and please relay any questions you 
might have;

-jason