G'Day Spencer, > Scheduling of the various NFS threads handling the > reads?
Yes - that fits - thanks; Here I've added a thread ID column, and this is NFSv4 which has the same jumping offset oddity: # ./offset.d TIME(us) TID CLIENT OP OFFSET BYTES PATHNAME 157083994915 543 192.168.1.109 R 0 4096 /export/share/bin/as 157084005459 544 192.168.1.109 R 4 28672 /export/share/bin/as 157084006244 543 192.168.1.109 R 32 32768 /export/share/bin/as 157084006508 543 192.168.1.109 R 96 32768 /export/share/bin/as 157084006864 543 192.168.1.109 R 128 32768 /export/share/bin/as 157084007201 544 192.168.1.109 R 64 32768 /export/share/bin/as 157084007521 544 192.168.1.109 R 160 32768 /export/share/bin/as 157084007973 545 192.168.1.109 R 192 32768 /export/share/bin/as 157084008155 546 192.168.1.109 R 224 32768 /export/share/bin/as 157084013581 547 192.168.1.109 R 256 32768 /export/share/bin/as 157084013717 544 192.168.1.109 R 288 32768 /export/share/bin/as 157084013917 545 192.168.1.109 R 320 32768 /export/share/bin/as 157084014009 546 192.168.1.109 R 216 4096 /export/share/bin/as 157084014430 543 192.168.1.109 R 240 4096 /export/share/bin/as 157084023525 546 192.168.1.109 R 352 4096 /export/share/bin/as 157084023871 543 192.168.1.109 R 356 28672 /export/share/bin/as 157084024595 544 192.168.1.109 R 384 32768 /export/share/bin/as 157084024848 544 192.168.1.109 R 416 20480 /export/share/bin/as Might be more clear if I add thread sched::: events to the output so we can see when which TID was blocked. Brendan This message posted from opensolaris.org