Rob, you got no credit because it’s all your fault in the first place.
Aaron, I used to think devtodo was the greatest thing ever. Until I discovered that a todo list is pretty much useless if you’re busy, because things on the todo don’t make it onto your schedule.
To ever get anything done, I need to actually schedule a time to work on a specific thing. I highly recommend it for any of you who never seem to get anything on your lists done.
Seemant, you know what bugs me the most about Bill Gates? He calls everything “rich.” What a bad pun for him to use that word. Anyway, the point is: the word “rich” is absolutely meaningless unless you’re talking about money. Which is exactly why I’m so disappointed when I see well-known Linux figures using it.
Corey, if you honestly think it’s a good idea for all the managers to invest in a copy of “The Five Dysfunctions of a Team,” perhaps you should consider putting some Foundation money toward it. Once we have some, of course. =)
I spent a few minutes installing and configuring NFS, TFTP and NTP, which were fairly trivial. The take-home points from NFS were:
- Set the daemons to use static ports, so your firewall doesn’t suck.
- Check out the mountpoint option, which prevents exports if an arbitrary location isn’t mounted. This stops the “underlying” directory from being exported if the mount didn’t work.
Later on, I might look into setting up NTP multicast to reduce the network traffic, if I have time.
After that, I started looking into the DHCP and DNS setup. ISC DHCP wouldn’t be so bad, but BIND looks to be a real bear to configure. Fortunately, Stuart pointed me to dnsmasq, a DNS server that can also act as a DHCP server. In particular, the dhcp-boot option may suit for my PXE setup.
Thanks to iggy (Brian Jackson), a Gentoo dev, I was pointed to missing support for one of the disk controllers as one of the problems. Once I got that in, the master booted like a charm.
In the meanwhile, I discovered that newer versions of gpm fixed a problem that made a rescue attempt without /usr pretty much useless, because libgpm was in /usr and a ton of stuff linked against it via ncurses.
I also switched from raidtools to mdadm on claims that it’s newer, better supported and more consistent.
The LiveCD’s RAID and LVM autodetection seemed broken for me, because I had to do everything manually. Perhaps I’ll get a chance to look into that later.
Still not sure about the nonpersistent error, but it would be nice if that resolved itself after the baselayout upgrade.
After some Googling, I’ve at least discovered that I’m allowed to have RAID 0 arrays with persistent superblocks.
From that page:
– RAID0 and linear now fully supports persistent superblocks ========================================
the bugs reported by Stephen Tweedie got fixed, and full support for persistent superblocks has been added. A RAID0 array from now on refuses to run if a device is missing or a numbering irregularity gets detected.
So, today I attempted to create the initial master installation. The disk setup is / and /boot on a 4G hard drive, and everything else goes on a RAID 0 of 2x120G drives. This should save me the effort of booting and having root on a RAID array.
In addition, to keep everything on the RAID on its own partition, I’ve set up LVM on top of the RAID rather than creating a large number of independent “arrays” out of individual partitions. This gives me more flexibility in partition resizing as well. Certainly EVMS could do similarly, but from my research, it seems like LVM is better-supported and more information is out there.
By the time I finished up the installation, it was about 6:30 p.m. and I had dinner plans for 7. So, I prayed and rebooted. But as you know, the first reboot _never_ works. I hit a couple of problems:
- Gentoo’s init scripts seem to start LVM before RAID. This generally doesn’t make any sense to me, because the reverse order is the one that would be useful.
- When it tried to start up the RAID, I got an error insisting that my RAID device had to be a nonpersistent RAID 0 or linear array. I didn’t see anything in the documentation about why RAID 0 had to be nonpersistent, so I made it persistent.
With a little luck and a little skill, I’ll recover from this tomorrow.
I’m beginning to wonder whether sending the hostname via DHCP makes any sense. We’ll still need DNS for name-based internode communication so we don’t need to deal with static /etc/hosts on every node or some other substandard solution. Maybe I’ll just forward the nodes to a DNS server and deal with hostname from there.
Quick poll: If you’re on a planet or read a planet, which one is it and how often does it update?
Seemant, if you think /var/lib/heimdal is more appropriate, why don’t you try to push it upstream instead of just deciding for us? If it’s a good idea, they’ll probably like it too.