We have noticed an interesting phenomenom on our Windows 2003 servers that are running Snapdrive for Windows. Every snapdrive operation adds a TCP connection to the filer at port 139 in a FIN_WAIT_2 status. These never clear until the server is rebooted. In the case of our heaviest used Oracle server, which has 4 LUNs mirrored to 3 different filers, this results in exhaustion of all available TCP ports over a period of time, resulting in no more CIFS authentication.
Has anyone else seen anything like this? Any ideas how to fix it, short of installing Service Pack 1 for 2003 as that is not approved for us to install yet?
Bug. And it's biting us BADLY.
http://now.netapp.com/NOW/cgi-bin/bol?Type=Detail&Display=234480
---snip--- ON TCP disconnect, CIFS client sends a FIN to the filer. Filer acknowledged the FIN and called soabort. Due to a software bug, filer didn't not send RST out to close down the TCP connection on the client, leaving the client socket in FIN_WAIT2. ---snip---
No published software workaround. It's rather nasty.
What Ontap are you running? We got pinched by it after upgrading from 7.2 to 7.2.1P1.