Hello util-linux-ng@:
I administrate a fair bit of systems with SCM software like cfengine, chef,
and puppet. When using these suites to administrate and provision
mountpoints, they all do the same trick to "UPSERT" mountpoints into
/etc/fstab: grep for an exisitng definition of a SCM controlled mountpoint
or device and replace the line if found, else append to the file a new
line defining the mountpoint plus commented garbage or no-op options to let
the admin/SCM know that this is a "managed" entry (in an "unmanaged" file).
I find this messy and increasingly unsustainable when managing mountpoints
for _many_ applications and users. Usually an SCM prefers that you
completely regenerate configuration files from a template rather than
incrementally mangling an "unmanaged" file. Since most SCM users don't
manage every single mountpoint in fstab in their SCM "policy",
/etc/fstab itself can't be generated from a template without risking
trashing SCM-unmanaged entries.
I'd like mount to support reading /etc/fstab.d/*. in addition to /etc/fstab.
Some files could be dedicated for SCM, or so-and-so user/program whilst
/etc/fstab remains the place for critical boot-time mountpoints that don't
ever need (automated) changing.
The cygwin environment's mount program is already using an /etc/fstab.d
directory in conjunction with a typical /etc/fstab file:
http://cygwin.com/cygwin-ug-net/using.html#mount-table
http://www.cygwin.org/cygwin/cygwin-ug-net/using-specialnames.html
Since altering libc to parse such a directory for getfsent(), getfsspec(),
and getfsfile() is out of the question (which is what I assume cygwin.dll
is doing), I was wondering if the mount program could instead get this
logic (in mount/fstab.c:read_fstab() I presume).
The directory could be walked as default behavior:
`mount /mnt/foo` or `mount -a`
- parse /etc/fstab for desired mountdev/point
- then parse /etc/fstab.d/* (nat-sorted)
or as demanded by an optionflag:
`mount --fstab=/etc/fstab.d/myDB /mnt/myDB` or `mount -a --fstab=/etc/fstab.d/SCM`
-only parse /etc/fstab.d/myDB
I see no need to keep fstab.d mounts out of /etc/mtab and breaking
all of the getmnt*() calls for such mounts.
Now there are automounter suites and that can provide such a tiered
configuration, but I find running an automounter daemon excessive (for such
a simple chore) and an unnecessary kernel<>userland performance drag for
highly-utilized local filesystems. I'd much rather use a simple one-time
initscript that will run mount with the appropriate options to iterate my
fstab.d files, or just `mount -a`.
An alternative to pathcing bin/mount is to run a wrapper script in place of
it that implements all fstab parsing and merely calls mount(). But
why keep that to myself and mangle every SCM I have to use such wrapper?
Thanks for reading
-joey
--
To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
