amd automounting daemon is unreliable

[Nathan Butcher]

I'm having some troublesome issues with filesystems mounted with amd in
FreeBSD 7.2-RELEASE

We have a backend file server wwith sharenfs attributes from ZFS
displaying NFS mountpoints to our front end machines:-

tank/export  mountpoint     /export                       local
tank/export  sharenfs       -maproot=0:0 server1 server2  local

and we have amd automounting directories on the backend file system

/etc/amd.d/amd.map.export:-

/defaults       type:=nfs;rhost:=wrfs;opts:=rw,vers=3,proto=udp,nodev
*               rfs=/export/files/${key}

/etc/rc.conf:-

amd_enable="YES"
amd_flags="-a /.amd_mnt -l syslog /files /etc/amd.d/amd.map.export"

There are a bunch of subdirectories in /files which get mounted across
the network when they are accessed on the front-end, and as a solution -
this has been working well.

...the PROBLEM is that it is unreliable. Sometimes amd hangs, and a
mounted directory - while listed in df becomes inaccessible and I get
"Permission Denied" errors while trying to access the mounted directory.
I'm using some monitoring on the amd process, but as it doesn't die to
trigger an alert, I have to wait until someone tells me that the
filesystem is rejecting them in order to respond tot he issue.

It also happens way too often for me to continue using this solution.
Unless I script amd to restart itself continuously in the event of a
crash... I cannot use it in a production environment.








-- 
フュージョン・ネットワークサービス株式会社 ■楽天ブロードバンド■
GOL事業部 システム技術部 ネットワーク・オペレションズ　グループ
Nathan Butcher / phone +81-50-5527-3611 / fax +81-3-3262-2224
https://secure3.gol.com/mod-pl/ols/index.cgi/?intr_id=F-2xsg6jS4u655