[Spread-users] Indications from spread

[John Lane Schultz]

Dutta Abhijit wrote:
> Can anyone help in this...Thanx in advance...
> 
> ~Abhijit 
> 
> -----Original Message-----
> From: Dutta Abhijit [mailto:adutta at intellinet-india.com] 
> Sent: Tuesday, July 04, 2006 2:57 PM
> To: 'spread-users-bounces at lists.spread.org'
> Cc: 'spread-users at lists.spread.org'
> Subject: FW: [Spread-users] Indications from spread
> 
> Thanx John for your help...but we have this observations..
> 
> When the cable from node1 is unplugged the event CAUSED_BY_NETWORK is
> received by node2. But no notification comes to node1 (ie. To self).
> 
> Note: node1 and node2 are in same group in different boxes with daemon
> running on each, connected through a LAN. 
> 
> So, is there a way by which we can make the daemon post a notification event
> to the application at node1 (to self) in case of cable pullout/network
> fails?
> 
> Also, if we kill the daemon in node1, the other (node2) gets the same
> notification (CAUSED_BY_NETWORK). So that means node2 can not distinguish
> whether it is a network fail or daemon crash. Am I right?
> 
> Your inputs...
> 
> ~Thanx
> Abhijit 
> 

Sorry for not responding yesterday because it was the USA's independence holiday.

Clients connecting through both nodes should have seen CAUSED_BY_NETWORK events 
reflecting that the "other side" of the partition had disconnected.  I'm not 
sure if/why both sides of the partition didn't see/report the disconnection.

Sometimes Spread will not (or is very slow to) *heal* a partition in a LAN 
unless client traffic is being sent, which everyone hears, but I don't think 
that is related to this issue with pulling the plug.

You are correct that Spread does not distinguish/report whether a daemon failure 
is due to a network partition or a daemon crashing.

Cheers!

-- 
John Schultz
Spread Concepts LLC
Phn: 443 838 2200
Fax: 301 560 8875