SUMMARY: NetApp File servers / Solaris 2.6 / Oracle 8i

From: Adam Bentley <Adam.Bentley_at_amdocs.com>
Date: Wed Apr 18 2001 - 07:53:57 EDT
First of all my thanks to the following :-

Jeff Kennedy,Dave Foster,Jay Lessert,Chris Miles

(apologies if I missed anyone...)

We've configured the links so they do not use autonegotiation which seems to
have helped. I am still seeing network services hang-up on the box which is
mounting the volumes from the NetApps  ( for any incoming network services
onto that box, telnet sqlnet etc.) but I believe this is an NFS issue -
because we are hard mounting the NetApps from this system it seems to
'block' until all the NFS traffic has been written. When you are doing a
database rebuild on the box mounting the netapp volumes, all seems fine, but
connections _into_ the box just hang until NFS has completed its work... I
guess we will have to live with this. Seems like more of a design issue than
a feature. :-(

Anyhow... 

Responses varied below :-

--------------------------------------------------------------------------

Don't use for Oracle


--------------------------------------------------------------------------
  
Here's the deal.  NetApp's serve Oracle data beautifully, the problem is 
  in the network (or the filer depending on how you look at it).  The 
  filer doesn't autonegotiate worth squat, make sure the port you're 
  plugged into is hard coded for 1000/FD. 

  For my own experience, I run several Oracle databases on filers and they 
  are great.  With that one caveat of course.  


--------------------------------------------------------------------------

Is this gigabit everywhere?  Solaris 2.6 gigabit drivers don't have the 
  best reputation (I've only run Solaris 8).  In particular, autonegotiation

  can be a problem, try forcing it off and doing manual setup on the
host(s), 
  switch(es) and servers. 



--------------------------------------------------------------------------

Hi, we have built a similar setup, except we are using Solaris 8.  Under 
  load testing I found that nfs mounts using udp would hang after a short 
  time of heavy traffic.  Switching to tcp shows no problems at all so far. 
  After talking to some people with extensive nfs knowledge it sound like
nfs 
  over tcp under modern operating systems is more advanced and stable than 
  nfs over udp, which we also concluded with our tests.  As far as 
  performance goes it is doubtful there would be much difference with 
  back-to-back connected gigabit connections which we are using. 
Received on Wed Apr 18 12:53:57 2001

This archive was generated by hypermail 2.1.8 : Wed Mar 23 2016 - 16:24:53 EDT