Hi, our t2k box running Generic_118833-36 experiences massive slowdown (upto 99% kernel cpu load) when launching several perl scripts. This might be related to memory starvation, but its hard to analyze (for me):
top reports 16G main memory, 800m free There are several zones running, containing a lot of big java processes (jboss). But these processes are not consuming much CPU (mpstat reports 99 idle for all CPUs/HWthreads). 3-5 perl process (http checks, light on the CPU) launch: box becomes unresponsive, smtx goes through the roof: CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 1 23 1 425 278 0 329 15 137 79483 0 48 6 93 0 1 2 32 0 586 216 0 305 13 115 83810 0 61 4 95 0 1 3 34 0 125 192 0 280 16 116 72529 0 46 9 90 0 1 4 28 0 206 236 0 490 37 154 63581 0 80 12 85 0 3 vmstat SR = 0 kernel CPU > 90% lockstat -s 5 -gkIW output: Profiling interrupt: 102240 events in 32.931 seconds (3105 events/sec) Count genr cuml rcnt nsec Hottest CPU+PIL Caller ------------------------------------------------------------------------------- 66840 65% ---- 0.00 3360 cpu[14] page_geti_contig_pages 66553 65% ---- 0.00 3360 cpu[14] page_get_contig_pages 66258 65% ---- 0.00 3374 cpu[14] page_trylock_contig_pages 51839 51% ---- 0.00 3404 cpu[5] page_trylock 32298 32% ---- 0.00 3430 cpu[5] mutex_vector_enter 29250 29% ---- 0.00 3315 cpu[0] page_get_freelist 19788 19% ---- 0.00 3112 cpu[18] thread_start [.....................] I can't make any sense from this. Any clues? Regards, Dirk This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
