SOS 3.5.1: Stops responding

[Marcus Ohlhaut]

After a major reconfiguration of our TCP/IP network structure, we frequently see the following issue on two of our SOS 3.5.1 servers:

They SOS service starts up and works fine for quite a while (approx 3-4 days), then all of a sudden, SOS clients can no longer connect. When this situation occurs, we observe the following:

- telnet to port 8888 opens a session, but we do not get the expected "200 Protocol Version: 2.0" lines.
- stopping and restaring the SOS service does not help, in this case the log file reports "Address in use"
- investigation with tcpview from sysinternals shows that a process named "system.exe" is listening on port 8888
- we need to reboot the server to recover

Any help will be appreciated.

Here is the start of the log file for reference
SourceOffSite Server 3.5.1
Professional Edition (Export Permitted)

Operating System: Windows NT 5.0

Java VM: Microsoft Corp. Version 1.1.4

Pin Support enabled.

Debugging output disabled.

VSS DLL synchronization enabled.

Number of licenses configured: xx

SourceSafe Initialization file(s) located at:
xxx\srcsafe.ini
xxx\srcsafe.ini

Listening for connections on unsecure port 8888


BTW: The log file contains records that list a computer name, others don't. On what does this depend?

Tue Apr 12 00:58:47 GMT+02:00 2005 - 10.2.8.186 (10.2.8.186) - Connection Established
Tue Apr 12 01:24:17 GMT+02:00 2005 - DGRAGG1 (10.2.8.58) - Connection Established

[lbauer]

Since SOS was working fine for quite a while, then something has changed on your network. There are issues with concurrency crashes caused by certain versions of VSS or by dual network cards:

http://support.sourcegear.com/viewtopic.php?t=1866

If none of these apply to your situation, the other possibility is that some other application is using port 8888 and causing a conflict with the SOS Server. This seems to be indicated by the "Address in use" error. I don't know what system.exe is, but you should investigate whether that app is the culprit.

You could try changing the port used by SOS to test this theory.

BTW: The log file contains records that list a computer name, others don't. On what does this depend?

Most likely you see the name in the log when SOS can resolve the host name from the IP address of the client. Otherwise, just the IP address is listed.