ZeroC

Archived

This forum has been archived. Please start a new discussion on GitHub.

Taking too much time to detect that the node is offline

Charles
Charles
edited June 2022 in Help Center

Hi,
I'm using Ice version 3.7.4. Registry and a node are deployed on different hosts. After the node status become active, I unplug the network cable of the node host.

For now, it takes over 10 even 30 seconds for registry to update the status of the node.

I have turned on all the trace configuration for registry including "Ice.Trace.Network" and "Ice.Trace.Protocol" and there is nothing between "Protocal: sending close connection" and a operation "nodeDown" during this no-response time.

I am not using Glacier2 and I have tried IceGrid.Regitry.NodeSessionTimeout property, and it does not work. I notice information of IceGrid Sessions in manual. Should I create session and involke keepalive on server?

Could you give me some advise? Thanks for your help!

Tagged:

Comments

  • Charles
    Charles

    IceGrid.Regitry.NodeSessionTimeout does help. By setting it as 2, the wait time is shorted to about 15 seconds, but it is still too long