Sun Solaris 2.6 vs. Sun Solaris 2.8

[twilk]

Have had a data loss problem with Sun Solaris 2.6 that is terribly
critical and very annoying. Sun Solaris 8 update seems to have exacerbated
the problem. We are using multicast, datagrams (UDP)...are confined
to using a 56 K line. Any ideas about why Sun Solaris 8 made the problem
worse?

Also, anyone with working knowledge of buffering and recovery algorithms and
available software that addresses this issue should feel free to speak up.

Thanks,
T. Wilk

 

[bigscouse]

Not sure about your problem.
You may be able to avoid the data loss by using TCP if that is an option.
Ged Jones

Top man

 

[twilk]

Unfortunately, my customer is very firm on staying with UDP, but thanks.
tw

 

[BVSJ]

Solaris 2.6 is having NFS version 2(NFS2) while Solaris 8 uses NFS version 3(NFS3).
When NFS was 1st developed in the '80s, UDP was chosen as transport protocol for LAN. Later in early '90s they developed a "transport independent" version of RPC which supported connection-oriented protocol like TCP/IP.

So, if you have missmatch on NFS version accross the client-server, you will have NFS data loss, network port collisions etc.

Try options like
-o proto=tcp or –o proto=udp
-o vers=2 or -o vers=3

Example :
mount -o vers=2,proto=udp \ nfsserver:/export/share1 /local_mount_point

Hope this helps.

 

[BVSJ]

Missed on prvious post.

man mount_nfs is very much discriptive on this on Solaris 8 as well as Solaris 9.