Agent service on Windows does not start

Question

Ok:
Grid Control 10.2.0.1, repository 10.1.0.4, Linux Red hat 4
Target node is a windows 2003 server (kinda, its on VMWare) with no service pack installed (don't ask). There is no oracle software on the target node except an oracle 10.2.0.3 agent. We just monitor the host up/down.
Ok this was working until the windows server crashed. It was restarted but then the agent would not start. It complained about not being able to load files and after three attempts it crashed. I removed it from grid, deinstalled the agent then reinstalled it with no errors. The same thing then happened, start the agent ok, then try an upload and it would fail and crash the agent:
2007-04-12 14:22:14 Thread-1820 ERROR upload: Failed to upload file A0000001.xml, ret =
-6
2007-04-12 14:22:14 Thread-1820 ERROR upload: 3 Failures in a row for A0000001.xml, we give up
2007-04-12 14:22:14 Thread-1820 ERROR upload: Error in uploadXMLFiles. Trying again in 72.00 seconds or earlier.
2007-04-12 14:22:16 Thread-2360 WARN upload: FxferSend: received fatal error in header from
repository: https://<grid_host>:1159/em/upload
FATAL_ERROR::500|ORA-20206: Target does not exist: Agent does not exist for
https://<target_host.com>:3872/emd/main/
ORA-06512: at "SYSMAN.EMD_LOADER", line 1700
ORA-06512: at line 1
2007-04-12 14:22:16 Thread-2360 ERROR upload: number of fatal error exceeds the limit 3
2007-04-12 14:22:16 Thread-2360 ERROR upload: agent will shutdown now
2007-04-12 14:22:16 Thread-2360 ERROR : Signalled to Exit with status 55
2007-04-12 14:22:16 Thread-3776 ERROR main: EMAgent abnormal terminating
Rebooted the box (its windows after all!!) but now its worse!
The agent service refuses to start at all, when I try to start it via the services panel I get straightaway 'Error 1053 The service did not respond in a timely fashion'. An emctl start agent from the command prompt produces a 'The service is not responding to the control
function'.
In the logs:
2007-04-13 09:03:32 Thread-1872 WARN http: snmehl_connect: connect failed to (<target_host.com>:3872): No connection could be made
because the target machine actively refused it.
(error = 10061)
2007-04-13 09:03:32 Thread-1872 ERROR main: nmectla_agentctl: Error connecting to
https://<target_host.com>:3872/emd/main/. Returning status code 1
But nothing is grabbing the 3872 port on the target host for the agent. The target host can 'see' the grid control server. There is no firewall in between.
I have an SR open, oracle are pointing the finger at a dodgy agent install (I've sent them the install logs to verify) or a firewall issue, but the target host certaintly does not have one.
I'm a bit of a loss now to be honest, any thoughts anyone? Not really used the agent on windows much.

Answer 1

If I clean out the logs and do:
emctl start agent
emctl status agent
The only logs produced are emctl.log:
4064 :: Fri Apr 13 09:38:28 2007::AgentLifeCycle.pm: Processing start agent
4064 :: Fri Apr 13 09:38:28 2007::AgentLifeCycle.pm: EMHOME is c:\app\ora\product\agent\agent10g
4064 :: Fri Apr 13 09:38:28 2007::AgentLifeCycle.pm:Exiting with 2
3972 :: Fri Apr 13 09:39:03 2007::AgentLifeCycle.pm: Processing status agent
3972 :: Fri Apr 13 09:39:03 2007::AgentStatus.pm:Processing status agent
3972 :: Fri Apr 13 09:39:05 2007::AgentStatus.pm:emdctl status returned 1
emdctl.log which is blank and emdctl.trc which is the output I posted earlier. Not very much to go on at all