Community-led development "The Apache Way"
If the problem is fatal to non-developers (e.g., any of the major daemons being down SMTP, HTTP, etc) send a note to noc@collab.net, which will act upon it as quickly as possible. If you don't get a response within 10-15 miniutes, as a backup call Brian Behlendorf on his cell phone #, which any ASF member can get for you.
If the problem is fatal only to developers (e.g., CVS commits are hosed, ssh access is screwy) please send a note to committers@apache.org. If it doesn't require root to fix, ideally someone can then leap in and fix it, with those with root ultimately responsible for fixing the problem.
Using "vnode" memory filesystem proposed by Tony Finch. this is now in
/etc/rc.local: vnconfig -e -s labels,reserve -S 200m vn0
disklabel -r -w vn0 auto
newfs -i 1024 /dev/vn0c
mount /dev/vn0c /tmp
chmod a+w /tmp
chmod +t /tmp
The -i 1024 to newfs is important, as cvs likes to create a lot of small
files.
Put "vfs.vmiodirenable=1" into /etc/rc.sysctl as a speed tweak.
Put "kern.timecounter.method=1" into /etc/rc.sysctl to address bind problem.
Make sure that fxp0 is set to "100baseTX <full-duplex>" by using "media 100baseTX mediaopt full-duplex" in the call to ifconfig.
much much more to come, I just needed a common place to start collecting anecdotal data.