Re: DropSafe Server Problems
People are having problems accessing my weblog and I’m not surprised; you see I rent server space on HostRocket who’ve generally been very good about support issues. At the moment what we’re faced with is the fact that someone else on the server is using MySQL as backing store for their webpages – or some other activity that presumably could not have been done by anything so mundane as a “filesystem” – and thus the load-average every few minutes spikes to some incredible number:
17:31:01 up 9:57, 3 users, load average: 23.49, 18.92, 22.75
17:31:18 up 9:57, 3 users, load average: 36.38, 22.10, 23.72
17:31:31 up 9:57, 3 users, load average: 43.29, 24.05, 24.33
17:31:44 up 9:57, 3 users, load average: 63.72, 30.37, 26.42
17:33:14 up 9:59, 2 users, load average: 81.13, 46.58, 32.61
…causing my (at least) piddling little CGI script to fail, because the process table is full; the logfiles are showing:
[error] (11)Resource temporarily unavailable: exec of cgi-bin/blosxom.cgi failed
…where “11” equates to EAGAIN, the Unix error code that I’ve always found most disappointing. Of course this only afflicts CGI content, the static stuff seems to be serving up pretty well.
HostRocket have installed a watchdog-style cronjob to monitor for this situation, but it tends to run into problems when the process table fills up; personally I’d slap a fair-share scheduler and process ulimits on all the users – in fact, going down the Solaris route I would put them all in Zones – but I am not in a position to suggest what HostRocket will be doing to control this in future.
This had been getting progressively more common since December, and it’s ironic that this peaks on the day that I get small links to my wesite from both Schneier and Slashdot.
Hohum. I’ve asked to be relocated to a different server, and will let you know what happens.