Broken Failover Logic in Livetribe SLP

[Nicholas White]

There is a bug in the SLP Server code that causes a problem...

TO REPRODUCE (using the standard directory server, and the client and
servers here: http://livetribe.codehaus.org/SLP-Examples-JMX

(1) Start the directory server, client and server. They should all
find each other
(2) Kill the directory server and the client
(3) Start the directory server, then the client
** the client will not find the server**

WHY

When the directory server restarts, it sends advertising messages.
However, the server doesn't respond as it thinks it has already
registered with that server. In the meantime, it advertises itself via
multicast. When the client is restarted, it finds the directory server
and interrogates it, so doesn't read the server's multicast messages.

CODE

In DirectoryAgentInfoCache, the following method is called when a
server receives a directory server advertisement:

   public void handle(DirectoryAgentInfo directoryAgent)
   {
       if (directoryAgent.isShuttingDown())
       {
           DirectoryAgentInfo removed = remove(directoryAgent.getKey());
           if (removed != null) notifyDirectoryAgentDied(directoryAgent);
       }
       else
       {
           DirectoryAgentInfo existing = add(directoryAgent);
           if (existing == null)
notifyDirectoryAgentBorn(directoryAgent); //*******REMOVE if(existing
== null)*******
       }
   }

In the above situation, existing is not null when the restarted
directory agent sends its advertisements, so the server never
responds. To fix this, remove 'if (existing == null)' so the server
always tells the directory agent what services it has.

Please let me know if you need any more clarification,

Thanks,

Nick

---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email

[Alan.Cabrera]

Thanks Nick.  Can you file a Jira issue?  Thanks!


Regards,
Alan

On Jan 20, 2009, at 1:37 PM, Nicholas White wrote:

> There is a bug in the SLP Server code that causes a problem...
>
> TO REPRODUCE (using the standard directory server, and the client and
> servers here: http://livetribe.codehaus.org/SLP-Examples-JMX
>
> (1) Start the directory server, client and server. They should all
> find each other
> (2) Kill the directory server and the client
> (3) Start the directory server, then the client
> ** the client will not find the server**
>
> WHY
>
> When the directory server restarts, it sends advertising messages.
> However, the server doesn't respond as it thinks it has already
> registered with that server. In the meantime, it advertises itself via
> multicast. When the client is restarted, it finds the directory server
> and interrogates it, so doesn't read the server's multicast messages.
>
> CODE
>
> In DirectoryAgentInfoCache, the following method is called when a
> server receives a directory server advertisement:
>
>   public void handle(DirectoryAgentInfo directoryAgent)
>   {
>       if (directoryAgent.isShuttingDown())
>       {
>           DirectoryAgentInfo removed =  
> remove(directoryAgent.getKey());
>           if (removed != null)  
> notifyDirectoryAgentDied(directoryAgent);
>       }
>       else
>       {
>           DirectoryAgentInfo existing = add(directoryAgent);
>           if (existing == null)
> notifyDirectoryAgentBorn(directoryAgent); //*******REMOVE if(existing
> == null)*******
>       }
>   }
>
> In the above situation, existing is not null when the restarted
> directory agent sends its advertisements, so the server never
> responds. To fix this, remove 'if (existing == null)' so the server
> always tells the directory agent what services it has.
>
> Please let me know if you need any more clarification,
>
> Thanks,
>
> Nick
>
> ---------------------------------------------------------------------
> To unsubscribe from this list, please visit:
>
>    http://xircles.codehaus.org/manage_email
>
>
>

---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email