Print

[jerome_tan]

When I try 2.5.1 with two drivers and two nodes on two machines.
Machine A, IP: 10.11.3.18
Machine B, IP: 10.11.3.16

I started two driver on the two machines with the same config file:
jppf.peers = server_1 server_2
 
# connection to server_1
jppf.peer.server_1.server.host = 10.11.3.16
class.peer.server_1.server.port = 11111
node.peer.server_1.server.port = 11113
 
# connection to server_2
jppf.peer.server_2.server.host = 10.11.3.18
class.peer.server_2.server.port = 11111
node.peer.server_2.server.port = 11113

jppf.discovery.enabled = false

The two driver starts properly and can find each other.

And I started two node on the two server, with configuration:
Machine A: use configuration: jppf.server.host = 10.11.3.18
Machine B: use configuration: jppf.server.host = 10.11.3.16

and "jppf.discovery.enabled = true" on both node.

The two node could find the two driver properly. But When I shut down the driver on Machine A, the node program can't switch the to Machine B automatically and keep printing:

Attempting connection to the class server at 10.11.3.18:11111
Attempting connection to the class server at 10.11.3.18:11111
Attempting connection to the class server at 10.11.3.18:11111
Attempting connection to the class server at 10.11.3.18:11111


But When I use JPPF2.5 with the same configuration on same server, the node program can switch to the driver on Machine B when the driver on Machine A is down.

Is it a bug?

[lolo]

Hello,

There is indeed a bug in the node failover mechanism in this configuration. I registered the following report:  3373217 - Node failover is not working

I need to investigate more on this issue, and will get back to you in this thread.

Could you let me know the value you set for the configuration property "reconnect.max.time" on each node? If it's not already done, could you try and set it to a low value, for instance 5, and let us know if this changes the behavior you observe?

Thanks,
-Laurent

[lolo]

Jerome,

I wanted to add that the basic failover machanism described here, will only work when discovery is enabled on both driver and node. If discovery is disabled on either, then the failover will not occur.
Also, the bug I mentioned previously has been fixed and the fix is part of JPPF 2.5.2, which was released yesterday.

Sincerely,
-Laurent