I have four web servers running two ip exclusive zones where nginx is the primary application. The systems have 16gb of ram, 1 L5630, on Intel S5500UR motherboards with onboard 82575EB Ethernet controllers. The load from the internet comes via round robin DNS. The four servers divvy up a high transaction load of very short lived connections. The outbound packet rate from the router to the four servers ranges between 15k and 60k packets per second. After one week of operation, response times talking to the back end (the only location which I measure) started creeping up from sub 1ms to peaks of 1000ms. I have isolated the problem to crossbow on the web tier. It seems that the only thing that makes the high response times go away is to halt the zones, delete the vnic, re-create it and boot the zones. Rebooting just the zone has no effect of the response times. Is there some kind of bit-rot in Crossbow? Is this a new bug? Thanks, j.
_______________________________________________ OpenIndiana-discuss mailing list [email protected] http://openindiana.org/mailman/listinfo/openindiana-discuss
