Thanks Ric for the info. I'll try to fill in the Linux autofs bit.
On Fri, 21 Jan 2005 [EMAIL PROTECTED] wrote: > [EMAIL PROTECTED] wrote: > I'm trying to integrate a linux box into a TRU64 UNIX network, and > am seeking some insight with automounting. Our TRU64 network uses > NIS to propagate data among our client systems. The systems have > been configured to use the NIS auto.master map to establish > automount links. > > Here's a snippet of the file: > > # Razor Links > /newrazor -rw,intr newrazor:/razor > /arazor -rw,intr cerberus:/razor > > # Misc. Alpha Links > /home -rw,intr mercury:/home > /vxdisk -rw,intr as2508:/vxdisk > > Is that from auto.master map or something like auto.direct? I.e., > auto.master often has a line like: > > % ypcat -k auto.master > /home auto.home -nosuid,hard,intr > /net -hosts -nosuid,hard,intr > /- auto.direct -nosuid,hard,intr > > Then auto.direct has something like: > > % ypcat -k auto.direct | grep release > /doclib/release_notes -rw,nosuid,soft server:/doclib/release_notes > > (Pay no attention to that soft mount flag. They are the work of the devil.) > > At startup, automount does _not_ create symbolic links. It _does_ create > "intercept points," NFS mounts that go to a NFS server within the automount > daemon: > > client 24% mount -e | grep non-ODE > client:(pid6035) on /usr/doc/public/non-ODE type nfs (v2, ...) > > When you access the intercept point, the mount is done and then the > symbolic link is created: > > client 25% file /usr/doc/public/non-ODE > /usr/doc/public/non-ODE: symbolic link to > /tmp_mnt/server/usr/doc/public/non-ODE > > client 26% mount -e | grep non-ODE > client:(pid6035) on /usr/doc/public/non-ODE type nfs (v2, ...) > server:/usr/doc/public/non-ODE on /tmp_mnt/server/usr/doc/public/non-ODE > type nfs (v3, ...) The Linux automounter doesn't use the intercept then symlink to triger mounts. The interception is done directly within the kernel so symlinking is not needed. > > I am already aware that the auto.master file format is incompatible > with autofs. I am working with automount2amd to generate a local > auto.master file for the linux box. The feature that I'm looking > for is the soft link generation upon autofs startup. > > That's generally considered a misfeature for various reasons and is > "corrected" in autofs implementations. Yes. Master map syntax needs work (sigh). > > In TRU64, > when automount is started, softlinks are created in root, based on > the NIS auto.master map. So, an ls -l in a TRU64 client would > display this: > > ... > lrwxrwxrwx 1 root system 20 Jan 17 16:46 arazor -> > /.automount/arazor > > No, per above. (The /.automount directory is due to overriding the /tmp_mnt > default.) > > My question is this: Is it possible to configure auto.master so > that when autofs starts up, These soft links will be automatically > created. When autofs is shutdown, the softlinks are automatically > removed. > > Oh, a Linux autofs question! Beats me. I'll yield the floor. :-) Don't think so - no. I've never seen anything in the code that would suggest that it would do this. It wouldn't be needed if the direct mount stuff was working. In fact, for single node mounts, like the ones you are having trouble with, the limited direct mount support won't work either. You will still need to manually creatre the symlinks. You'll need to wait until I get to release a 4.2.0 alpha before you will even be able to test it. Don't ask when. I don't know. All I can say is that, fixing the direct mount limitations is my personal itch. So it's right up top in priority and work on the alpha will be starting straight after 4.1.4 is released (4.1.4 beta1 is due this week). Ian _______________________________________________ autofs mailing list [email protected] http://linux.kernel.org/mailman/listinfo/autofs
