On Wed, Dec 22, 2010 at 08:52:30AM +0000, Dietmar Maurer wrote: > Corosync v1.3.0 (single node) > Debian Squeeze AMD64 with latest 2.6.32 kernel > > When I run "corosync-pload" it prints: > > # corosync-pload > Init result 1 > > The process never stops (but I can stop it with cntrl-c), but it seems to > work anyways: > > Dec 22 09:32:46 maui corosync[2409]: [PLOAD ] 1500000 Writes 300 bytes per > write 2.495 seconds runtime, 601307.250 TP/S, 172.035 MB/S. > Dec 22 09:32:53 maui corosync[2409]: [PLOAD ] 1500000 Writes 300 bytes per > write 3.062 seconds runtime, 489821.674 TP/S, 140.139 MB/S. > Dec 22 09:33:01 maui corosync[2409]: [PLOAD ] 1500000 Writes 300 bytes per > write 4.372 seconds runtime, 343112.460 TP/S, 98.165 MB/S. > Dec 22 09:33:09 maui corosync[2409]: [PLOAD ] 1500000 Writes 300 bytes per > write 4.369 seconds runtime, 343358.870 TP/S, 98.236 MB/S. > Dec 22 09:33:53 maui corosync[2409]: [PLOAD ] 1500000 Writes 300 bytes per > write 3.475 seconds runtime, 431594.847 TP/S, 123.480 MB/S. > > If I now start cpgbench I get: > > /corosync-1.3.0/test# ./cpgbench > 463802 messages received 1000 bytes per write 10.000 Seconds runtime > 46380.121 TP/s 46.380 MB/s. > 470350 messages received 2000 bytes per write 10.000 Seconds runtime > 47034.864 TP/s 94.070 MB/s. > 460633 messages received 3000 bytes per write 10.000 Seconds runtime > 46063.231 TP/s 138.190 MB/s. > 443571 messages received 4000 bytes per write 10.000 Seconds runtime > 44357.016 TP/s 177.428 MB/s. > > Everything OK, but if I also start corosync-pload I get a corosync crash: > /corosync-1.3.0/test# ./cpgbench > ... > cpg dispatch returned error 2 > > and the syslog shows: > > Dec 22 09:39:45 maui corosync[2409]: [PLOAD ] 1500000 Writes 300 bytes per > write 2.184 seconds runtime, 686771.055 TP/S, 196.487 MB/S. > Dec 22 09:40:03 maui dlm_controld[2479]: cluster is down, exiting > Dec 22 09:40:03 maui fenced[2464]: cluster is down, exiting > Dec 22 09:40:05 maui kernel: dlm: closing connection to node 3 > > Can someone reproduce that? How can I further debug that?
Hi Dietmar I pload is not destrustive, but it does not play nicely. It is a tool to compare sending messages without IPC been involved (compare with cpgbench not to be used at the same time). Some issues with pload: 1] In service/pload.c you can see it doesn't ask for totem queue space like the IPC request code does in main.c, it just hammers away at totem_mcast(). ( It doesn't call totempg_groups_joined_reserve() ) So when I run these two apps I get: corosync: cpg.c:1449: message_handler_req_lib_cpg_mcast: Assertion `result == 0' failed. Which to me means that cpg reserved totem queue space but it was not honored (pload used it up). This will also starve other applications from sending as they check for totem queue space. 2] message_handler_req_exec_pload_start() It doesn't send a response to the client (why you needed to CTRL-C) -Angus > > - Dietmar > _______________________________________________ > Openais mailing list > [email protected] > https://lists.linux-foundation.org/mailman/listinfo/openais _______________________________________________ Openais mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/openais
