> It seems the memory is going to the kernel.
I have this problem on my web proxy tier, but not on my backend systems. I have checked kernel core dumps for memory leaks, switched the Ethernet cards/drivers, tested exclusive versus shared interfaces on the zones, and tweaked everything I could think of. I haven't found a solution. I run the same motherboard every where and I only have this issue on the proxy tier. I put 96gb of ram in one box and watched the system consume it all. Ultimately, I increased the RAM across the board to push the reboot cycle out a couple of weeks. My longer term plan is to develop the software stack on Linux and migrate the proxy tier to CentOS. It sickens me that we can't get this right, but if Linux is the best tool for the job then that's what I will use. Fortunately, the data tier doesn't suffer from this problem and that's where we really need the data handling capabilities of ZFS. My hardware platform on the web tier is Intel SR2625URLX, LSI 9211-8i, Crucial M4s for storage. The primary application running is nginx. I have nearly no writes to disk. We have such a fire hose of http traffic that we cannot possibly write logs on the proxy tier. By the way, it is nice to find you. James Carlson insisted I was the only one having this problem ;-) j. root@www003:~# for i in `seq 1 5`; do date; echo ::memstat |mdb -k |egrep "Kernel|MB"; sleep 120; done October 15, 2012 03:32:37 PM PDT Page Summary Pages MB %Tot Kernel 3659602 14295 44% October 15, 2012 03:34:39 PM PDT Page Summary Pages MB %Tot Kernel 3660353 14298 44% October 15, 2012 03:36:41 PM PDT Page Summary Pages MB %Tot Kernel 3661008 14300 44% October 15, 2012 03:38:43 PM PDT Page Summary Pages MB %Tot Kernel 3661708 14303 44% October 15, 2012 03:40:45 PM PDT Page Summary Pages MB %Tot Kernel 3662362 14306 44%
_______________________________________________ OpenIndiana-discuss mailing list OpenIndiana-discuss@openindiana.org http://openindiana.org/mailman/listinfo/openindiana-discuss