View Full Version : Transparent NFS

19th December 2010, 03:04 AM
Is there a way to make a user's 'home' file on a client computer default to the shared 'home' file on a server? I know this can be accomplished with a network login, but I'm not sure I want/need to go through the hassle of setting up remote log-ins for a simple home network.

Case: catastrophic hardware failure resulted in lots of lost pictures.
Result: pissed off wife
Solution: home nfs. NFS housed on Raid5 with frequent back-ups and periodic, off-site back-up storage.

I would like to set things up so that when my wife logs into her account on the laptop, her 'home' account defaults to her NFS folder.

Any help is appreciated.

19th December 2010, 08:31 PM

You can do it in several ways, and network login is not required.

1) Cheapest and dirtiest - write down the NFS path in /etc/fstab and it will be mounted at boot time.
2) A better one, IMHO - use pam_mount which will mount it on-demand when you log in.

Don't forget that depending on the underlying filesystem type, you'll probably need to take care to match local UIDs and remote ones.


20th December 2010, 12:43 AM
I agree with Assen, You could easily place an nfs mount at /home or /home/swmbo.
You can (and perhaps should) mount on top of a non-empty directory. So for example if you want everyone to have their default home dir on the NFS, then you can leave all the /home/username/.. stuff in place and also have a copy on the NFS share, then mount the NFS at /home. The advantage is that if the NFS mount (at boot) fails for any reason, then at least everyone can still login to their stub home directories.

The SunMS "yellow pages" commands yip, ypcat, ypasswd and such were all invented to support networked NFS home directories and distributed passwords - and this is still supported on Fedora, but it's an old, crufty, somewhat insecure and hassle-ful way to do things. I do not advise that approach.


I have even more concern about your explanation

Case: catastrophic hardware failure resulted in lots of lost pictures
OK - so there was no duplicate copy, no backup and presumably the disk was destroyed otherwise there are loads of data recovery Linux tools, even pix recovery tools (search this forum). We've all been there. Thankfully I've never lost more than a days work due to this sort of unpleasantness, but it can and does happen for a lot of humans-are-imperfect reasons.

Result: pissed off wife
My experience is that pissed -off-wife is usually a natural disposition, and not a results of a specific event. It's difficult to make the causal chain, from pix lost to PO'ed SO, despite that that is the content of her recent tirade. She didn't marry the prince and live happily ever after in the castle, is more likely the cause of her angst. You can't fix that one - don't try.

Solution: home nfs. NFS housed on Raid5 with frequent back-ups and periodic, off-site back-up storage

You bring up 3 distinct issues:

Realize there are two categorical types of data loss possible; media-loss and user-loss. The media may fail (so we have RAID1, RAID5, and backup-copy schemes). Also the user may accidentally delete files and not realize it till later. RAIDn and single copy duplicates of file-systems with missing files is not a solution for accidentally deleted files. We need historical backups, incremental or otherwise.

NFS per se doesn't prevent file system losses from either sort of data loss. I'm not sure why you include it. Still this is a great resource for NAS,

RAID5 means you need to have 3 (or more) live disks in the server and your get 2/3rd (or 1-1/n) of the total space. It's tolerant to a single disk media-failure. The reduces one form of media failure. The common PSU voltage spiking is still a SinglePoint-of-Failure(SPoF). Your house burning down is a SPoF. If you store a backup across town, then there are still regional disaster SPoF from flood or armed attack. Thunderdome is planetary SPoF, Sun going white dwarf is a solar systemic SPoF. You have to make some realistic decisions about how important your record are after nuclear EMP destroys all electronics, or after a Katrina wipes entire your city. So sure - you know in advance that tax records, wedding pix, biz records are important enough to send to a data storage service, but IMO those generic tourist pix of Cancun or the Bahamas are just fine sitting on a pair of disks on the same PSU. Make some intelligent decisions about the level of failure protection needed for each group of records.

RAID>0 is little more than a dynamically sync'ed backup and this does nothing or accidental deletions. RAID5 is also an expensive thing to keep up 24x7.

So there are a lot of complex trade-off here, and your goals and needs are unclear.

If you can accept a 24hr roll-back copy, then it's a lot cheaper to keep a separate backup disk idle in the system and have it perform backups at midnight. Of course you could perform incremental backups at hourly intervals or whatever if that's better for your needs. As oppose to RAID5, you'll invest a little more in disk space (but it's very cheap) but you won't be paying $100/yr to power a NAS 24x7 and you won't be wearing down your commercial (non-enterprise) drives to the point of a 20 month service interval. My ballpark is that a home RAID5 NAS 24x7 will cost ~$250/yr amortized and a spare 1.5TB backup drive just spun-up for backups will cost more like $15/yr.

If you are concerned about the SPoF, you can use a NAS system with WOL or timed wakeup to perform backup then go back to sleep. But if you are performing backup-copy to the NAS then you already have the original, making the NAS RAID5 data redundancy rather excessive IMO.