LTC Owner is: ankigarg.com LTC Originator is: ankigarg.com While trying to duplicate bug #26701 (a LTC internal bug, an NFS problem on the 2.6.16 RT kernel not recreatable on 2.6.17), found that NFS copy of huge files does not complete on RHEL5 RT. This is a priority issue and would mostly be fixed once the IRQ priority values are finalized on RHEL5 RT. So, keeping the priority low till set_kthread_prio values are decided upon. Will test it once that is done. This can recreated as follows: Requires 3 systems on RHEL5 RT. On system A, nfs mount a directory structure on B & C. From, A, nfs copy huge files, recursively, from B to C. The IRQ priority values in set_kthread_prio-1.0-1.x86_64.rpm need to be adjusted to correct this issue. Discussion on this is happening on rhel-rt-external list. Mark Salter from RedHat has worked on this.
----- Additional Comments From ankigarg.com (prefers email at ankita.com) 2007-05-07 00:29 EDT ------- By making changes to the priority values in the rtgroups file, part of set_kthread_prio implementation, the NFS copies completed successfully. Attaching a patch to the rtgroups file.
Created attachment 154238 [details] rtgroups_nfs_safe.patch
----- Additional Comments From ankigarg.com (prefers email at ankita.com) 2007-05-07 00:31 EDT ------- fix nfs related daemon priorities Adjusting the priorities of the NFS related daemons, rpciod, lockd and nfsd, ensures that these are not starved due to other rt threads. This would enable the nfs related operations to complete.
changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |FIXEDAWAITINGTEST Resolution| |FIX_BY_IBM ------- Additional Comments From ankigarg.com (prefers email at ankita.com) 2007-05-14 05:32 EDT ------- Marking this as FIXED_AWAITING_TEST so that RedHat can start on the process of testing the suggested changes in the set_kthread_prio.conf (rtctl) values.