ZBX_TCP_READ() failed

[kc5kse]

Ever since upgrading from 1.4.x to 1.6, I get the following, sporadically, from the master log:

1280:20080922:102431 Timeout while answering request
1280:20080922:102431 NODE 1: Error while receiving answer from Node [2] error: ZBX_TCP_READ() failed [Interrupted system call]


Our Networking department swears up and down nothing has changed and it's the upgrade of Zabbix that is causing the problem. What can I do to pin this down one way or the other?

[nelsonab]

Usually the message means the server has killed the connection to the client before any data was received, or before the server was able to understand the data received. Telnet to the agent manually and test the key to see if it works.

You should see your item result and an EOF tag, I forget what the exact tag is but it's easily recognized.

[kc5kse]

This occurs during communication between Master and Subordinate nodes. I only see it from the Master's log. And it occurs sporadically. I'll see this, then sometime later there will be a mass SEND/RECEIVE as they sync up.

[kc5kse]

I am able to telnet to each host from the other on ports 10050 and 10051. However, a tcpdump from the master doesn't show anything while a tcpdump from the subordinate node gives me tons of info when I grep the master's ip.

Since I am not the network admin, can some kind soul please point me in the right direction here? I hate to go back to 1.4.6 but I may have to.