How Efficient/Stable are the am-utils?

Steve Baum asks: "I'm thinking of replacing a current tangle of NFS cross-mounted disks with the am-utils system, which maintains a cache of mounted filesystems that are demand-mounted when first referenced and unmounted after a period of inactivity. I was wondering if anyone had used this system in a moderately large (40-50 disks on 10-15 machines) environment and, if so, how efficient and stable they'd found it to be."

Comments

Not very good

[Anonymous Coward]

I used to use them on my combined network of Sun Ultra2s, Linux-based PCs, and the occasional older SGI. We tried to switch to am-utils to cut down on the network connections and excess traffic created by buggy NFS implementations (eg, those found in Linux kernel-based implementations (the two user-land implementations we later tried were fine, i fogget what either of them is called))

At any rate, am-utils failed to properly unmount shares on the SGI and Linux machines, which lead to threadlocks on Ultrix and file-descriptor shortages (followed shortly by VM crashes). Needless to say, this was dropped, and we switched to a combination of WebDAV and shared RAID arrays (distributed clustering and NAS systems were also evaluated and later dropped from consideration, mostly due to cost).

Advice to you? am-utils is not all it seems. Dont forget that NFS is not the only way to get remote file access. Re-evaluate what you need. Sometimes, things as simple as GNOME VFS or Emacs's efs are enough. Otherwise, look at the things I mentioned above.

-rick

Re:AFS anyone

[O]

Hell yeah
OpenAFS baby!