Hey,
Last night I moved apache temporarily to deathray, installed Linux from
a DVD on to the spare disk on Murphy, moved the web server there and ran
some tests.
In short, it worked perfectly, zero hitches. There was an issue with
networking, but it turned out to be me noobing and configuring the wrong
ehternet ports, it worked fine once I told it to use eth2 and 3, not 0
and 1.
I installed apache with suphp, suexec, pubcookie, etc, all configured
and running all of the vhosts, worked fine from a user point of view
(even wiki logins worked, they never work). Tested a few sites (static
ones like my own, PHP/database ones like gamessoc's, and some things
under ~testwww that usually break badly configured servers), all worked.
It was under the regular load of 2am (admittedly not much), plus 15
instances of a script on a few external servers making randomised
requests to various pages drove load to about 12 with no noticeable side
effects.
Booted back into Solaris after about half an hour of this. The bugs on
OpenBSD showed up after 5/10 minutes of apache being switched on, so we
can assume Linux doesn't have the same problem.
I'd suggest we switch Linux back on permanently next weekend. We can
leave it on the spare disk for a few days (maybe a week), that gives us
a window to find bugs, etc, and if there's anything serious we can jsut
reboot back into Solaris at any time. After a week or so is up, assuming
nothing goes wrong, we could then transfer Linux onto the main disks in a
software raid 1 setup.
A side effect of all this is that we now have a cold spare web server on
Deathray :) If murphy catches fire or anything we can just configure the
WWW IPs on deathray, send a gratuitous ARP and fire up it's apache to
minimise downtime while we figure out what's wrong with
$main_web_server.
-Andrew
