 |
 |
Hi Winnie, I have enquired about this (on LCG-GOLLOUT, I believe) a few months back when I upgraded our WN tarball. The response was that wn- list.conf should include a single line that is the headnode hostname, the actual machine on which the tarball is being configured. Cheers, Gianfranco On 21 Sep 2009, at 15:23, Winnie Lacesso wrote: > Dear All, > > On our HPC of which PP/LCG gets a portion, the WN are busy running > jobs. > So we run yaim for the HPC WN tarball install on the HPC headnode, as > people are supposed to compile etc etc on headnode, not on WN. > > We were at 3.1.24-0 & are getting ready 3.1.33-0. > Yaim seems to have broken now as it expects the node on which yaim is > run to be a WN. But this is nonsensical for HPC; the WN are busy > running jobs, so running yaim on the WN (where gpfs is having a very > hard time, so let's not add to it) would interfere with user jobs! > > ERROR: The WN you are configuring is not defined in the WN_LIST file > /exports/gpfs/gridpp-shared/lcg/yaim-conf/wn-list.conf > ERROR: Configuration error ! > ERROR: Configuration error ! > lcg@bluecrystal1> > > That's true, bluecrystal1 = HPC head node, it's not a WN & is not in > wn-list.conf. > This has never been a problem in previous working yaim versions. > > What is the solution - fake that the HPC headnode is a WN in wn- > list.conf, > or fix yaim, or interfere with user jobs by running yaim on a busy WN? > > Grateful for advice -- Dr. Gianfranco Sciacca Tel: +44 (0)20 7679 3044 Dept of Physics and Astronomy Internal: 33044 University College London D15 - Physics Building London WC1E 6BT