incubator-wadi-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jules Gosnell <>
Subject Re: activemq 4 - snapshot & peer
Date Thu, 05 Jan 2006 12:04:42 GMT
Bill Dudney wrote:

> Hi All,
> Good news on the amq front. Whatever was going wrong with the peer://  
> protocol not being able to grow after shrinking to one node is fixed.  
> I have successfully created a cluster of 3 nodes, killed off 2 of  
> them and regrown it to 3 and the new group is equally able to process  
> messages.
> Jules - there were two other issues we were hoping were fixed in the  
> 4.0 code, could you point me to the other two?

sorry it has taken a whlie to get back to this.

the issues were (as I recall)

1). shrinking/growing issue - hopefully resolved - excellent :-)
2). lazy and slow allocation of inter-peer connections wreaks havoc by 
causing timeouts in big clusters
3). restarting a node within a heartbeat may confuse AC as a node may 
not disappear, but its connection may die (speculative)

(2) is the real show-stopper, but you have to build a large cluster (I 
was trying with 40 nodes or so) in order to get to thescale at which it 
happens... Hiram suggested that connections were made preemptively in 
4.0, so I am hoping that this is fixed.

(3) may not be a problem, unless we lengthen the heartbeat rate, which 
you might do again in a large cluster. I have heard reports of this 
issue, but not experienced it at firsthand (I think).

Thanks for looking at this, Bill - its critical stuff.


> Thanks,
> Bill Dudney
> MyFaces -
> Wadi -

"Open Source is a self-assembling organism. You dangle a piece of
string into a super-saturated solution and a whole operating-system
crystallises out around it."

 * Jules Gosnell
 * Partner
 * Core Developers Network (Europe)
 * Open Source Training & Support.

View raw message