Marella P-G19460 wrote:
> Hans, 
> 
> The purpose is to stall (MDS based) communication, with a node that has
> exited cluster, as soon as possible.

So AVD-AVND heartbeating discovers that a node has left before TIPC, is 
that correct?

BOM xml contains:
>   <NCSScalars>
>       <sndHbInt>1000</sndHbInt>
>       <rcvHbInt>3000</rcvHbInt>
>       <clusterInitTime>40000</clusterInitTime>
>   </NCSScalars>

These configuration values determine the characteristics of the AVD-AVD 
and AVD-AVND heartbeats?

Regards,
Hans

P.S. An idea, AVD-AVND heartbeating could be removed and instead OpenSAF 
could use the TIPC topology service?


> 
> -----------
> Example:
> Assume a system with nodes named A,B,C and D.  If the node D were to
> exit the system, then "AVNDs" on A,B and C would get a message saying
> "node D has left the cluster". This would cause invocation of
> "tipc_reset.h" in A,B and C nodes to disconnect them from D. The
> expected result is that the TIPC links towards D (viz A<->D, B<->D,
> C<->D) get immediately disabled. 
> -----------
> 
> This feature is useful if the TIPC link tolerance (== link health-check
> timeout) is set to value higher than the default 1.5s. A higher TIPC
> link tolerance avoids unnecessary TIPC connection failures in scenarios
> where a temporary link failure (and its resolution) takes greater than
> 1.5s. Due to lack of a better link reset/disable support in TIPC,
> "tipc_reset.sh" uses a workaround of temporarily bringing down the link
> tolerance to a low value.
> 
> Hope this helps.
> 
> Regards, 
> Phani
> 
> PS: AVND = Availability Node Director (part of SCAP/PCAP executables)
> 
> -----Original Message-----
> From: [EMAIL PROTECTED] 
> [mailto:[EMAIL PROTECTED] On Behalf Of Hans Feldt
> Sent: Monday, November 05, 2007 6:12 PM
> To: [email protected]
> Subject: [Users] Purpose of tipc_reset.sh?
> 
> I currently don't have tipc_reset.sh working since tipc-config 
> is not in 
> /opt/TIPC (and it will never be). But what is the purpose of this 
> script? I can see it is used by avnd when a node has left the cluster.
> 
> Regards,
> Hans
> _______________________________________________
> Users mailing list
> [email protected]
> http://list.opensaf.org/maillist/listinfo/users
> 

_______________________________________________
Users mailing list
[email protected]
http://list.opensaf.org/maillist/listinfo/users

Reply via email to