Note: Since the web won't be up when you need this document, print out a new copy whenever you change it and put it on the rack...
From time to time we get power outages and someone has to go around and power-up the DiRT machines. With the FreeBSD machines doing things like sharing a file server and using other FreeBSD machines as routers it is useful to power up the machines in a specific order: Servers, routers, and finally end-stations behind the routers. Nothing extermemly bad will happen if you do these in the wrong order but, obviously, you won't be able to do much that's useful until they router and servers are up. And, you'll probably have to do some reboots. The following order is recommended for the FreeBSD machines:
buzzard (facilities does this) ----> wait for buzzard (because it's a file-server for everything) tyagi ----> wait for tyagi (because it hosts /usr/dirt) moorthy, dirt, michigan, oberon, othello ----> these are general purpose file servers goldberg, topsecret, titus, romeo, rosalind... ----> all machines in 137 (NFS dependencies are complex, hence please check your production machine after a power outage) ----> wait for topsecret and goldberg (connect our nets to the dept.) daffy, elmer, bollella, borg, chen ----> wait for the machines above the rest of the exp network and test and development network machines.
Basically, the order is file-servers, routers connecting our nets to the department, then routers internal to our network, then the rest of the machines. Most of the production machines could actually be in that last group. Bringing machines up earlier just means you will have to reboot them all to get the proper file-systems mounted. (And it makes it more difficult to do the reboot with our automated tools because file-systems aren't mounted...)
Use this script to reboot sets of machines. e.g.
/usr/dirt/bin/reboot_machines -M /usr/dirt/config/machines.expnet.norouters
The following machines will not restart by themselves: thelmalou, speedy and oberon (?)