NFS soft versus hard mounts

We had an “interesting” debate in the office yesterday. And when I say interesting it was more like “heated”. Soft vs. hard NFS mounts is really something that gets people going, mainly because some people has been annoyed by a process or server hanging when its NFS server has become un-available.

Soft mounting does have its uses, but it is only really safe on a proper read-only mounted NFS drive. If you do want to soft mount on a read-write drive, then, to ensure data consistency, your application must be programmed to handle un-expected situations which almost no programs using NFS drives are.

Hard mounts are always the recommended way to go! Throw in “intr” if you will, which should help allowing you to kill off hung processes, even though the Linux brigade seems to have removed the intr option for some reason.

If you do have a hung NFS mount, your main priority should be to fix the NFS server, or get the people in charge of it to fix it, and as soon as the NFS server is back, your processes will spring back to life and most importantly, you will not have suffered any data loss.

If you really do think soft mounts are the way to go, do you do this on your SAN drives as well? It’s probably not even possible and for good reason.

So, to re-iterate, never ever use soft mounting unless you really know what you are doing and have taken the appropriate precautions. (If you’re reading this post for advice on whether to use soft or hard, then you most certainly need hard!)

This entry was posted in AIX and tagged , , , , . Bookmark the permalink.

Leave a Reply