[ale] Wow! Check out these load averages!

Dylan Northrup docx at io.com
Tue Mar 11 13:49:15 EST 2003


A long time ago, (11.03.03), in a galaxy far, far away, Michael D. Hirsch wrote:

:=On Tuesday 11 March 2003 09:26 am, Jonathan Glass wrote:
:=> Turns out someone had shutdown a machine that this box attaches to via
:=> NFS, so the box was hanging waiting for the NFS connection.  It wouldn't
:=> reboot for the same reason.
:=
:=A trick I've learned is to either mount the NFS drives with the "soft" 
:=option, or with "hard,intr".  In either case, this should allows 
:=operations to fail and exit when the server goes down.
:=
:=For some reason, the docs say that the latter is recommended.  I had always 
:=used the former with no bad effects.  Check the docs for the "intr" option 
:=as I'm not sure I spelled it right.

If you're running something where data integrity is crucial (database,
e-mail, mp3 servers :-) and you don't want to worry about whether your
applications deal gracefully when writes don't return what they're supposed
to, doing a hard mount is better.  With a soft mount, the I/O can fall on
the floor and lead to data corruption if your application doesn't know how
to deal with it.  Using the 'intr' option does allow you or your application
to respond to interrupts to kill a process that's hung while waiting for
response from an unresponsive NFS hard mount.  If you specify 'hard' without
specifying 'intr' you can get into the situation where your load skyrockets
because of unresponsive programs stuck in I/O wait.

Hope this helps.

-- 
Dylan Northrup <*> docx at io.com <*> http://www.io.com/~docx/
"Harder to work, harder to strive, hard to be glad to be alive, but it's 
 really worth it if you give it a try." -- Cowboy Mouth, 'Easy'

_______________________________________________
Ale mailing list
Ale at ale.org
http://www.ale.org/mailman/listinfo/ale






More information about the Ale mailing list