Despite of the dragon names, they are two bits. I had a very good initial experience with the system deployment tool cobbler but the romance lasted only a little. Unfortunately on a research center, we constantly update the systems. Since I mounted my cobbler on one of our systems, it needs a reboot from time to time. And there is where the problem occurs. Even if I take care of scheduling the services at booting time, even if everything coming mounted as it should, cobbler doesn’t show up. Fortunately, I know ho to fix it: systemclt cobblerd restart works usually fine.
The other problem with cobbler is the distribution & kickstart. Both are very sensible: if you make a modification on the system, or on the kickstart file, you will probably end up with a non-bootable node. So do backups, my friend! Later on, I may post our kickstart.
And what is the problem with Leginon? Appart from the fact that you probably don’t use it, of course. The main issue is on the multiple configuration files. So let’s say you want to connect one microscope with your leginon server, and then another different one. If you don’t know where your configuration is, and where your calibration, you may end up running a Krios like a Spirit, or viceversa. And the information on the configuration files is not “transparently” linked with a gargantuan database. You can get an introduction on Leginon here if you are interested.