topic GlobalProtect Reporting Port Exhaustion on Win 10? in GlobalProtect Discussions

GlobalProtect Reporting Port Exhaustion on Win 10?

Brett.Hoshaw — Fri, 04 Sep 2020 17:18:11 GMT

Anyone else seeing a lot of this lately? I've been researching this over and over again for about 2 weeks and all signs point to Windows being the culrpit, but the only time we ever see this happen is when users are connect to GlobalProtect.

Here's the culprit log from GPS:

connect failed with error 10055(An operation on a socket could not be performed because the system lacked sufficient buffer space or because a queue was full.)

Things I can verify:

Reboot fixes (winsock reset also fixes it, but requires a reboot anyways)
- Sometimes 2
Users don't lose active sessions
Users can no longer browse anything via DNS
New connections can be made via telnet
Internet access is still available (ping and traceroute show that affected devices are traversing public hops and can reach them)
Local network is also affected
Port exhaustion has not actually occurred
- One user had 80 of 16,000+ ports in use (clearly not exhausted)

Re: GlobalProtect Reporting Port Exhaustion on Win 10?

Max.Segura — Sat, 05 Sep 2020 01:54:51 GMT

That's interesting. What version of Windows and build are you on? Also what does netstat show on the end user machine?

Re: GlobalProtect Reporting Port Exhaustion on Win 10?

Brett.Hoshaw — Tue, 08 Sep 2020 21:41:33 GMT

Netstat shows that only 80 ports were in use (out of 16,000+ total - checked that via the CLI to make sure it wasn't something small).

Windows builds are ~~9309~~ all around 1909 18369.959

The only common denominator seems to be GlobalProtect.

I saw a new error today which resulted in a disconnect with the same symptoms:

Bio Enum Databases failed. hr = 0x0

What still intrigues me is the fact that it seems like affected machines just seem to lose DNS. Public hops works, everything seems to be functioning except no browsing (DNS related). All signs point away from GPC, but none of our non-vpn users are having this issue and they remotely connect in different ways.