Franco Broi wrote:> Is it normal to have so many ldlm process consuming so much CPU? I have > just one process writing to the lustre filesystem via NFS.No, not at all normal. It looks like they may be spinning on something. While this is happening, can you do a SysRq-t and email the resulting output from /var/log/messages? A few more questions: Which version of Lustre? Are all services running on this node, or just some? Are there any messages on the console? Anything special to reproduce it? Just start Lustre, export NFS, and run ''cp''? Thanks-- -Phil
Is it normal to have so many ldlm process consuming so much CPU? I have just one process writing to the lustre filesystem via NFS. PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 3691 root 10 0 0 0 0 SW 21.1 0.0 4:36 0 ldlm_bl_31 3680 root 9 0 0 0 0 SW 18.7 0.0 4:33 1 ldlm_bl_20 3670 root 9 0 0 0 0 SW 16.9 0.0 4:32 1 ldlm_bl_10 3677 root 9 0 0 0 0 SW 15.9 0.0 4:34 1 ldlm_bl_17 3675 root 9 0 0 0 0 SW 14.7 0.0 4:29 1 ldlm_bl_15 3682 root 11 0 0 0 0 SW 14.3 0.0 4:41 1 ldlm_bl_22 3672 root 9 0 0 0 0 SW 13.5 0.0 4:39 0 ldlm_bl_12 3664 root 9 0 0 0 0 SW 8.3 0.0 4:14 1 ldlm_bl_04 3679 root 9 0 0 0 0 SW 8.1 0.0 4:51 0 ldlm_bl_19 3668 root 18 0 0 0 0 RW 5.9 0.0 4:43 0 ldlm_bl_08 6077 root 9 0 0 0 0 SW 3.3 0.0 30:57 1 nfsd 3685 root 9 0 0 0 0 SW 3.1 0.0 4:14 1 ldlm_bl_25 3689 root 19 0 0 0 0 RW 2.9 0.0 4:48 1 ldlm_bl_29 3663 root 9 0 0 0 0 SW 2.1 0.0 4:17 1 ldlm_bl_03 3684 root 9 0 0 0 0 SW 2.1 0.0 4:31 0 ldlm_bl_24