|Well, I'm impressed... (and curious, too)
||[Feb. 28th, 2002|01:33 pm]
I downloaded the LiveJournal code from CVS two days ago and spent the majority of yesterday tinkering with it. It took a bit to get everything going, but finally I got pretty much everything working. (Displaying journals doesn't work, theres currently an issue with it, but its already being discussed in another thread). Went to sleep last night, woke up today the webserver had crashed and was spewing kernel messages at me: '/kernel: swap_pager_getswapspace: failed' Started at 3:46 AM, and while the server kept running, it did try to sync and reboot itself. When I caught it at 5:40AM, the console would still respond, but was lagged ~45 seconds (well, obviously, no swap space). It failed its sync at ~5:46, and I force rebooted it then. I was not able to get a process snapshot before the system went down.
After rebooting, I've let the system run all day. (1:40PM, now) when I took a look at the process table. 280 sleeping perls, 88 zombied perls, all the physical memory used, almost all the swap space used - had about 100mb of swap left. I killed all the perls, shut down apache, and removed my cron entries, just to keep myself sane. All my swap space freed right up, along with most of my physical memory.
So, I've started to wonder... is this normal, or has the code I've grabbed (and cvs update'd a few times since my first co) been bugged?
For reference, the system I was testing on is a Dual PPro 200 with 192mb of physical ram, another gig in swap space, and the database resides on a different host (my database server). Running FreeBSD 4.5-STABLE, updated as of less than 2 weeks ago. Apache 1.3.23, with mod_fastcgi and mod_php4 loaded. mod_ssl is compiled, but for the test was starting with 'start' and not 'startssl'.
Any thoughts are appreciated. Thanks. =)