On Tue, Jul 14, 2009 at 04:39:19PM -0600, Robert Thurlow wrote: > Joerg Schilling wrote: > > >>>Do you see a benefit from replacing technology that is known since more > >>>than 20 years by something that does not yet have even test users? > > >Which benefits? > > Caveat - this is about a follow-on case. > > I'm glad you asked. The current automounter hasn't had a decent > amount of work on it for years now, and it has a lot of issues, > strictly as a piece of code in a Solaris system. Maps are read > [...]
It's not that the code is unmaintained -- these are architectural flaws in the automounter concept. Since the automounter is _client_-driven the namespace information must be distributed to all clients, which works well enough in a small environment, but does not scale to the Internet. Any automount map distribution protocol will just not lend itself to having clients dynamically adjust as the namespace changes -- clients would have to poll too often and detect changes, or the protocols would have to support exchanging diffs, and so on and on. The very notion of automount maps that need to be distributed also tends to centralization of namespace management, which presents its own scalability issues: every time you zfs create/destroy you must edit an automount map on some server, or do some LDAP creates/modifies? ugh, ugh!. Mirror mounts already solve problems that the automounter can't possibly. Referrals extend that solution past a single host. That is good. Nico --
