Bug #29835 | Agent dup uuid errors occur due to network issues | ||
---|---|---|---|
Submitted: | 16 Jul 2007 21:23 | Modified: | 22 Jul 2007 21:42 |
Reporter: | Bill Weber | Email Updates: | |
Status: | Closed | Impact on me: | |
Category: | MySQL Enterprise Monitor: Server | Severity: | S1 (Critical) |
Version: | 1.2.0.6610 | OS: | Any |
Assigned to: | Darren Oldag | CPU Architecture: | Any |
[16 Jul 2007 21:23]
Bill Weber
[18 Jul 2007 1:32]
Darren Oldag
Made session handling more resilient to network timeouts. A session allocation requires an 'acknowledgement' of the session back from the agent before it is considered assigned. Then, normal agent/dashboard communications can occur. If the session times out, packets from the agent are dropped until a new session is ACK'd. Two ACKd allocated sessions are not allowed for the same agent uuid. --------------- Note, a good way to test this would be to yank a network cord for two minutes, enable a firewall, etc, just so both services stay alive. Let a network timeout occur, then undo the nasty network (re-enable it). Then, see if everyone catches back up.
[18 Jul 2007 15:50]
Darren Oldag
not in the 1.2 branch yet, but committed to trunk for review.
[19 Jul 2007 21:02]
Darren Oldag
pushed into the last build.
[22 Jul 2007 21:42]
Andrew Cwik
fixed in 1.2.0.6744 To test this I've had 57 agents running for two days and none of them have stopped prematurely. The agent log files and catalina.out file don't have any entries referencing a duplicate uuid.