Skip to Content.
Sympa Menu

sm-discuss - Re: [SM-Discuss] [off-topic] Admin advice

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: Thomas Orgis <thomas-forum AT orgis.org>
  • To: sm-discuss AT lists.ibiblio.org
  • Subject: Re: [SM-Discuss] [off-topic] Admin advice
  • Date: Mon, 16 Apr 2007 22:58:12 +0200

Am Mon, 16 Apr 2007 14:12:02 -0700
schrieb Eric Sandall <eric AT sandall.us>:

> You may want to make sure you have the latest stable kernel, latest
> nfs-utils,
> and latest portmap userland installed and running. While I do not have a
> lot
> of mount requests at home, I do transfer a lot of data (gigabytes per day,
> somedays) via NFS.

Heh, I just have to do that: Gigabytes per day? You call that a lot?
$ ifconfig
RX bytes:3426805790089 (3268056.6 Mb) TX bytes:7223145501531
(6888528.3 Mb)
$ uptime
11:15pm up 98 days 8:02, 1 user, load average: 1.27, 1.37, 1.31

So even if you just take the mean, that's 75.6GiB a day;-)
But actually this box has been sleepy for many days and when there is
work, it's more like 50MB/s permanent for days. NFS seems to be happy
there. It's a 2.6.18.2 like kernel (an OpenSuSE install). And I really
hope it's continuing to be happy since being a NFS server for the SATA
RAID is the main function of that box.

On the other hand... we have distributed home directories, with stuff
cross-mounted (automount) over the workstations. That actually works
for the linux workstations... but the old eMail / NIS server with Tru64
didn't really like that; it had to mount a user's home from somewhere
to check for forward config and the like for every mail.

> have to restart the network service as all NFS traffic stops, but other
> traffic continues.

I once got some corrupted files via NFS, but that was hardly NFS's
fault: The hardware RAID driver crashed the kernel, apparently after
taking in some packets for writing the files. Clients waited for all
the downtime (hard mounts) and when we brought the server up again
continued... assuming that some packets have been written that were
actually not. I had some range with zeroes instead. But the kernel
crashed while writing to the hardware in the driver... I guess not only
NFS would be affected there.

>
> NFS has it's problems, but it is the most simple networked filesystem to
> setup. If you want more robustness, there's AFS or Coda which have failover
> and all kinds of useful tools. ;)

What I'd like to have is some proper authentification (without
sacrificing any performance, of course;-). NFS volumes are an open,
writable book for anyone who has root on one of the workstations... and
if you have PC workstations, every user is potentially root (unless you
have some good protection in bios setup, perhaps).

AFS looks so overly complicated. sshfs is an easy one, but I doubt I'll
want to use it to transport all the data over gigabit links. The server
will have some load with all the encryption. Though it would be the
secure thing.


Thomas.

Attachment: signature.asc
Description: PGP signature




Archive powered by MHonArc 2.6.24.

Top of Page