On Tue, 2017-05-02 at 12:17 -0400, Jamie Jackson wrote: > On Mon, May 1, 2017 at 8:58 PM, Ian Kent <raven@xxxxxxxxxx> wrote: > > On Mon, 2017-05-01 at 16:02 -0400, Jamie Jackson wrote: > > > I'm new to Autofs, but I've had one auto-mount going for a few weeks > > > now. It came time for me to add a second one, but I ran into "ignoring > > > duplicate indirect mount /mnt/non-icf/my" > > > > All you need to understand is the concept that there is one master map that > > contains top level automount managed mount points and entries in it have a > > specific format. > > > > The master map associates mount maps with these top level automount mount > > points. > > > > The maps specified in master map entries typically have more than one entry > > (although if one entry is all you need that's fine too). > > > > > > > > I see from https://www.spinics.net/lists/autofs/msg00000.html that I'm > > > supposed to use a different syntax to get both of these mount points > > > to coexist: > > > > > > /mnt/non-icf/my/secureftp > > > /mnt/non-icf/my/testsftp > > > > > > I'm reading https://www.centos.org/docs/5/html/Deployment_Guide-en-US/s1-n > > > fs-c > > > lient-config-autofs.html > > > , but I'm having trouble absorbing the content. > > > > > > Could someone help me convert what I've got to what I need? > > > > > > https://gist.github.com/jamiejackson/460ba0a6526d6251554a157821f526da > > > > First lets look at the master map. > > > > If you're not using entries that are present in the default comment them > > out. > > That would be /misc, /net and +auto.master. > > > > Commenting out the last one is actually important because, if you're not > > using > > the automount entry of nsswitch, the sources present in the default install > > (this configuration file isn't owned by autofs) can cause some error message > > noise in the log and introduce an unnecessary delay at startup. > > Great, I see the difference after having removed it. > > > Now these two entries: > > /mnt/non-icf/my /etc/auto.my_sftp_test \ > > uid=492,gid=488,--timeout=60,--ghost,--debug > > /mnt/non-icf/my /etc/auto.my_sftp_prod \ > > uid=492,gid=488,--timeout=60,--ghost,--debug > > > > The uid and gid options are not used by autofs. > > When I had them defined, the files showed as having the local > user/group that I wanted. After having removed them, they show as the > uid/gid of the remote user. I'm not sure how to reconcile my findings > with what you said, above. Could you elaborate on that? Umm .. now I think about it what I said is plain wrong (almost). It's been a while since I've had to think about this but I think master map entry options that aren't autofs specific will be used as global options (appended to map entry options by default) for mounts so it probably comes down to what fuse will do with the options. Without a mounts having been triggered for these map entries the options should have no effect. > > > Personally I always use the autofs configuration to enable debug logging. > > I take that to mean that i should remove the `--debug` option from the > master map, and add `LOGGING=debug` to auto.master. (Done.) Mmm .. oldish syntax .. what version, OS? It shouldn't matter though. > > > Either way you need to ensure that syslog is capturing log level debug or > > greater for the log entries to be sent to syslog. > > > > If you're using systemd journalctl to instect the logs then setting log > > level > > doesn't matter but if there is a lot of logging systemd will start dropping > > entries. > > > > Also I prefer to use the browse option rather than --ghost. > > It looks like `BROWSE_MODE=yes` is the default, so I just removed the > `--ghost` flag. It depends on distribution. The internal program default is browse (because other widely used autofs implementations had this default) but in RHEL, for example, the default installed configuration sets browse_mode to no because that was the expected behaviour of autofs before browsing was implemented. And there can be some unexpected semantic behaviours when using it. But that also depends on kernel version, it gets hard to describe at this point so I'll not say more about it. snip ... > > > Beware that if you try to set debugging for the fuse mounts themselves (ie. > > not > > handled by autofs) I think the sshfs mount process won't go into the > > background > > causing the mount to never complete and that will result in unexpected > > behaviour. > > Yup, I learned that the hard way in > https://bugzilla.redhat.com/show_bug.cgi?id=1434600. Thanks for that. Right, I thought the name looked familiar, ;) Ian -- To unsubscribe from this list: send the line "unsubscribe autofs" in
