Home Help Center


mxiaoaimxiaoai Member Xiaorong MaOrganization: PersonalProject: Cluster Schedule Simulator

I wrote a simple Java application for the control of dispatching tasks using IceGrid(3.5.1). It runs at master node. Then my servers which do real task run at node1 node. They are distributed at different physical nodes.

If there is a new task, it will be hand out to the proxy for an allocatable object registered with the given type. Then master records the identity. After that if the connection between the nodes is broken, master will first resume the proxy using identity. If the communication cannot be recovered, master will release the object. Afterwards master goes on to allocate the object by type again.

I just did one test case: after the master hand out a task successfully, I turned off node1. Due to long time spent on task processing, before the end of the task the master will reallocate it. At first, Ice.NoEndpointException was reported about 10 times, because I deploy 10 servers at node1. Then IceGrid.AllocationTimeoutException came out. Later I turned on the node1, but master cannot get the proxy at all. It's IceGrid.AllocationTimeoutException all over.

So what's wrong with the master? If I restart the master, it will be OK.

Any ideas?
Thank you!
Sign In or Register to comment.