The only slight issue I have with mosh is that it does not support cursor shape. I use this a lot, both in my vi-mode shell and in vi.

mine is that it does not seem an active project

find that too critical for something like that

it'll probably reactivate when some cryptographic advancement considered important comes out

until then, it's in a maintenance mode

it works just fine otherwise

I did see occasional commits on git

Is there an automounter expert in the house?

I'm trying to convert something which I could do with /etc/fstab, but that would leave it mounted all the time (and this should be lazy-mounted).

geometry: AFAIK, there is no "lazy" mounting in FreeBSD. What are you trying to accomplish?

Trying to do a single mount using the automounter (which is effectively lazy mounting).

What automounter?

Like fuse or something?

The automounter (/etc/auto_master).

automountd running?

Yup, the difficulty is figuring out the right syntax for /etc/auto_master.

The portion of the man page under "MAP SYNTAX" seems to be the right section, but I'm left puzzled at WTF "key" should be (some portion of the path I'm guessing, but I'm puzzled).

geometry: What are you trying to mount?

And what's currently in /etc/auto_master?

An NFS mount, some of the options are specific to the particular mount and /net doesn't work well.

I've only used automount a couple times in the past. IIRC, I used it to point to a separate file that specified mount types and locations?

Like, /etc/automnt_dirs or something.

geometry: Can you provide the exact line in /etc/auto_master?

I've tried several things none of which give me indications of success, I know what I would use for `mount` or /etc/fstab, but I'm puzzled about converting to automounter.

I believe the "key" is just the name of the mount (and perhaps what the mount name becomes if available). So, something like: "media -fstype=nfs,rw nfs-server:/share-path" should work?

Or: "movies -fstype=nfs,ro nfs-server:/movies"

Something like that?

"too many arguments at /etc/auto_master, line 8"

What's line 8?

<mountpoint> auto -fstype=nfs,<other_opts> <nfsserver>:/<nfspoint>

Actually that is line 9, line 8 begins with '#' so I thing it is counting from 0.

How are you specifying <mountpoint>? Also, I don't think you need "auto" in there.

Leading "/" in the <mountpoint>?

"test -fstype=nfs,whatever nfs-server:/share" should mount as /master/test or just /test?

Since it is a test, it is /tmp/test_point and wouldn't "auto" be the key?

No. The key is the first column entry.

What I did was something like "/- auto_mount" in the /etc/auto_master file. Then, in the /etc/auto_mount file, I'd add something like:

"/mnt/nfs-share -fstype=nfs,rw nfs-server:/nfs-share"

topcat001, Just noting that mosh is going to use the terminal so cursor shape would depend on the terminal not mosh.

And I could add additional mountpoints and shares to the /etc/auto_mount file.

It's been a long time since I have set up the automounter. But for key think trigger point to trigger the mount happening.

Yep. It's the "auto" point.

For NFS, maybe I didn't use the -fstype= arg. Maybe I did just use the "-intr,nfs,rw" instead? Looking at the auto_master man page, that would make sense, I suppose.

I don't recall ever seeing -fstype= and it was just immediately nfs mount options.

Problem is I need options besides intr.

"failed to execute /etc/autofs/special-intr,<other options>".

List them out as comma separated options.

I recall there being three types of maps. Host maps, direct maps, indirect maps.

Indirect maps more of a rule to generate the paths, are the hardest to get a mental understanding of, so best to start with one of the other two first while trying to understand them.

This is only for 2 mount points, so I doubt indirect maps would be worthwhile.

The host map is perhaps the simplest. Any access to /net/foo will run "showmount -e foo" and see what directories are exported for nfs mount, and then mount them as /net/foo (where foo is a hostname).

Then direct maps are pretty much what it says on the label. It maps this to that directly and mounts it.

A direct map seems plausible for this situation.

Another tidbit I remember is that it is good to always use a subdirectory of / or deeper. Otherwise doing something like "ls -l /" will cause everything to automount. Better to have it /nfs/foo instead so that this only would happen if someone "ls -l /nfs/" and that is less of an issue.

Yep. I really should mess with automount again. I did have it working and it worked just fine AFAIR.

It did take some finagling, though.

Used it industrially when I worked for HP in the lab but haven't had a case since then with so many mount points that doing other than just hard mounting in /etc/fstab was more reasonable.

Same. I just add to /etc/fstab and done.

Granted, late,bg,retrycnt= are needed in most cases, but I digress.

When there are potentially a thousand mount points only sporadically needed then the automouter is a practical solution. This appears in the compute farms where jobs are sent with qsub or other and any random compute server might mount some few mount points to have the job run on it. More practical than having all compute servers mount all thousand mount points.

I would prefer /etc/fstab, but this is a setup I've inherited and things are a bit problematic for always being mounted.

If I were home I would walk over to the bookcase and pull out my O'Reilly book on NFS and NIS and it would get me back up to speed. But I am not home and that book is not available to me at this time.

So here is an example that /I think/ should work: /mnt/projectfoo -intr,rw,async other.example.com:/srv/projectfoo

Pretty sure. I can't test it. But if not that then it would be pretty close to that.

Yeah. I remember it being pretty simple once I got it working. It kind of "clicked" like, "Ah, yes! Okay."

Hint: Don't use "soft" with nfs. Using soft mounts can lead to silent data corruption. Don't do it.

Yup.

Other trick is use a private network or ULA for NFS and your local services.

Okay, I may have gotten things working.

Woot!

I think the example would likely be much better if it had "/- auto_map" and a sample /etc/auto_map file.

Having "-noauto" really doesn't offer any useful clue.

geometry: It does have that example. "/- auto_example" and then and example of "/etc/auto_example" with the mount points.

You can use "auto_WHATEVER" AFAIK.

Yes, but a default installation has /etc/auto_master with the line "#/- -noauto" which isn't too helpful.

Ah! I gotcha'. Yeah, that seems fair.

What was your line that eventually worked?

That example line has probably been in that file since 1990.

Mostly simply going for a direct map and discarding my preconception of the file containing actual mount points.

No example, huh? :)

I would think direct map for NFS would be the best choice.

I fear the example might merely be 20 years old instead of 35 years old, but /home and /var/mail would be classic mount points.

Back in the day it would have been /usr/mail then. And depending upon the system /users instead of /home. It's a System V thing.

/users was indeed a SysV thing, though I think they accepted /home by 2000 or so.

Oh, I didn't mean specifics of mount points. Just the syntax of the /etc/auto_map or whatever you ended up using. Just for prosperity's sake in case someone else ends up looking for the same info in the future.

I think most universities running BSD used /u and then a hierarchy below in order to handle a large number of accounts.

I think you need a copy of the O'Reilly NFS and NIS book. It's the missing documentation.

For the universities we've recently (in the last 10 years or so?) set up, they use /home or /usr/home. It's fine until they expect it to exist locally when they're logging into a non-NFS mounted /home node or something.

Sometimes there was a heirarchy below on the server, but desktops simply mounted them directly on /home (or /u) since desktops didn't need that.

ant-x I like to use root everywhere, I live dangerously like that yes.

We use /mnt/home for NFS home and /home for local users. to be fair we have almost zero local users, everything lives in NFS and LDAP.

I want to use spiped because it's really neat, but I haven't found a relevant use case for it for myself

If anyone is in the business of updating the handbook, it teaches about the old bsdlabel thing which is deprecated. That section must get updated with gpart

I've got some FreeBSD VPSes. IPv6 networking stops working after a while after reboot. ping6 fe80::1 still works. But anything else doesn't. Hints?

netstat -rn6 : termbin.com/hljq3

okay... somehow the loopback interface has claimed fe80::1 which is the ISP gateway...

THIEF!

okay now it works. Somehow, two redundant addresses have been added both to primary ethernet interface and the loopback one

oops

How can I find out which process is responsible for doing so?

I am reporting a bug. What component should I choose?

I guess it's either kern or misc

tsoome: Pressing esc and entering got get doesn't do anything. It outputs something, but I can't see it.