Products | Versions |
---|---|
TIBCO SmartSockets | - |
Not Applicable | - |
Resolution:
To address the issue of "ghost RTservers" an RTserver will abort a PRIME-in-progress if it receives updates to the cloud state after registering a new connection (and sending it a PRIME) but before completing it and bringing it up as a full server. This may be seen when launching RTservers and RTclients at the same time. Recommend launching RTservers first and then RTclients.
Specifically, in SmartSockets 6.8.0 the following was introduced:
Resolved an issue that could lead to various problems including "ghost" servers. If the cloud's subject routing state changes while two servers are priming, the priming server would not be notified of the change, and would end up in an inconsistent state. Now (beginning with version 6.8.0), if something changes during a prime, the prime is aborted by the server that detects the change. This is seen by the other end as the server connecting and immediately disconnecting. On the next server_reconnect_interval the affected RTservers will reinitiate the connection attempt. This change does not affect compatibility with old servers which will handle being disconnected in the same manner, however older servers are not able to detect the priming problems and therefore cannot initiate the corrective action themselves.