Best practice for using multiple GSA PS installs?
@Sven curious to hear your thoughts on this:
I have a ton of proxies that are in GSA PS, using various sources, Proxy Buddy, scraping a ton of my own sources. Sometimes I can have 30k proxies that are tested and working by GSA PS. The beautiful thing is that Scrapebox test shows 80% of these SOCKS proxies don't work, but with GSA PS they work like a charm. It's amazing really.
The problem is that, when I use the same list across multiple PCs/servers, I imagine the proxies are getting hit at the same time, and being detected as down even though they are not. For example, I just checked a server that was running 800 threads between 4x installs, it had only 100 valid proxies left!
Edit:
I think I found a solution Sven. Could you add a feature that is "Automatically re-test 'Not Working' Proxies after X minutes?" I just did a re-test of not working proxies and it worked perfectly, now I have 26k again, although going down of as they're being used. I'm pretty sure this feature would solve the problem...
Edit 2:
Uhh... so after I re-tested the Not Working proxies, I check back and within 1 minute it went from 26,000 -> 14,000...?
Wow, after re-testing Not Working, the proxies almost INSTANTLY went to 0 alive... It has to be that I'm using the same list between 2 servers or something... because these proxies are up when tested by GSA PS, but then detected as down instantly when used for SB.
Edit 3:
What does "Test error: No buffer space available" mean? I was looking into this last night and tried some stuff in Windows 7 like increasing the max ports etc... hmm well I think the timeouts are too high, trying to change them now. Just downloaded TCPView and like 5 minutes after not running into connections, there were thousands open.
Maybe I should install this but I don't have a "Multiprocessor PC" so I didn't install it:
https://support.microsoft.com/en-us/help/2577795/kernel-sockets-leak-on-a-multiprocessor-computer-that-is-running-windo
But, does GSA PS create "loopback sockets"?
I have a ton of proxies that are in GSA PS, using various sources, Proxy Buddy, scraping a ton of my own sources. Sometimes I can have 30k proxies that are tested and working by GSA PS. The beautiful thing is that Scrapebox test shows 80% of these SOCKS proxies don't work, but with GSA PS they work like a charm. It's amazing really.
The problem is that, when I use the same list across multiple PCs/servers, I imagine the proxies are getting hit at the same time, and being detected as down even though they are not. For example, I just checked a server that was running 800 threads between 4x installs, it had only 100 valid proxies left!
Edit:
I think I found a solution Sven. Could you add a feature that is "Automatically re-test 'Not Working' Proxies after X minutes?" I just did a re-test of not working proxies and it worked perfectly, now I have 26k again, although going down of as they're being used. I'm pretty sure this feature would solve the problem...
Edit 2:
Uhh... so after I re-tested the Not Working proxies, I check back and within 1 minute it went from 26,000 -> 14,000...?
Wow, after re-testing Not Working, the proxies almost INSTANTLY went to 0 alive... It has to be that I'm using the same list between 2 servers or something... because these proxies are up when tested by GSA PS, but then detected as down instantly when used for SB.
Edit 3:
What does "Test error: No buffer space available" mean? I was looking into this last night and tried some stuff in Windows 7 like increasing the max ports etc... hmm well I think the timeouts are too high, trying to change them now. Just downloaded TCPView and like 5 minutes after not running into connections, there were thousands open.
Maybe I should install this but I don't have a "Multiprocessor PC" so I didn't install it:
https://support.microsoft.com/en-us/help/2577795/kernel-sockets-leak-on-a-multiprocessor-computer-that-is-running-windo
But, does GSA PS create "loopback sockets"?
Comments
I'm thinking that when the buffer space ran out, GSA PS just basically started canceling the proxies at a rapid rate, because it had no choice.
For anyone else who has this problem: Here are the registry settings:
HKLM > SYSTEM > CurrentControlSet > services > Tcpip > Parameters
TcpTimedWaitDelay - 30 (decimal)
MaxUserPort - 65534 (decimal)
TcpTimedWaitDelay cannot be set lower than 30. Apparently if you try it, it will default to 120-240.
The MaxUserPort number I used is also the highest the OS will allow. I’ve had no problems out of these settings yet.