-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Aryeh Gregor:
All the commands for managing system startup, controlling services, dealing with the filesystem and disks, debugging, etc., etc. are totally different between Linux and Solaris. [...] You also have to install all software twice, once for all the Solaris machines and once for nightshade. And keep track of two streams of OS and software updates, and install them separately.
Two other reasons:
1, our entire infrastructure is set up for Solaris, since we have about 12 Solaris servers, and 1 Linux server. This means we have automated installation profiles for Solaris, management software (like Puppet) configured for Solaris, etc. This infrastructure makes it much easier to manage the Solaris systems, but there's no equivalent for nightshade. It would be possible to set it up, but that would be even more effort, and for only a single server. Having all servers use the same infrastructure is much easier.
1.1, as a side effect of this infrastructure, the Solaris systems are standardised, because everything is automatic. If we want to add a new Solaris login server, I can just create a 10-line text file describing the server (its IP address, the fact that it's a login server, etc.), start a network install, and half an hour later, we have a new login server configured identically to the existing ones. If we ever needed to reinstall nightshade with Linux, the new system would probably be different, since it's difficult to exactly replicate one system from another.
2, having one Linux server prevents us from using some Solaris features, e.g. role accounts for MMTs (like the old stable server had), or RBAC for access control, because these features don't interoperate with Linux. This is only a minor issue, but it is annoying at times.
- river.