BIG problem, 90 users 24/7 !!!

[ekkas2]

I had this problem in January, my Wingate server just freeze. Black screen, nothing, not even mouse. There is nothing wrong with the server hardware.
In January Adrien had a look at our server and adjusted MTU which seemed to fix the problem, well, we've been running just fine until I had to re-install XP recently on the PC because of corrupt files that scandisk couldn't fix.

PLEASE HELP URGENTLY.

I've set the MTUs as it was and recently increased network buffers. I have a saved copy of the registry settings, but other settings like advanced WG settings and NIC windows settings are not in there.

We have about 90 users on a wireless network and if the server freeze I have to drive there, reboot, delete history files reboot again and it works for a few hours. This happens at random and is VERY VERY annoying.

We are at the point where we will have to ditch WG if we can't sort it out soon. We have already lost 3 customers because they 'can't work with such an unreliable service'....

Is there anyone out there that can give me some idea what can be done, what should I look at? In January, just as 6.2.1 was released Adrien sent me exe's that I had to copy over the current ones which I do not have anymore. This problem also did not exist prior to 6.2.0 and started since then, up until January when Adrien fixed it with his magic and now it is worst than even since a re-install.

Ekkas

[adrien]

Hi

I think we did something else as well to solve the problem to do with your configuration, not just MTUs.

That would have been lost with the reinstall.

I also think you had a DNS loop set up due to the WinGate server being part of an active directory.

Adrien

[adrien]

I seem to recall you set in the past the MTU onto the network adapters themselves in the OS using DrTCP, rather than just in GateKeeper. did you use that this time?

I found the old emails. You had a serious problem with your router and/or ISP breaking Path MTU discovery. I take it you didn't get anywhere with them trying to resolve that?

[adrien]

Hi

I found your email and logged in with VNC. Looks from the registry like the MTU settings were only set in WinGate.

Problem with this is then the OS doesn't know about the MTU, so local outbound traffic from the server can have problems.

You'll need to set the MTU on the adapters in the registry or using DrTCP.

and reboot.

Cheers

Adrien

[ekkas2]

Thank a lot for the reply.

1) We do not use AD
2) I did use TCP optimizer to set MTUs but I'll use DrTCP as well...
3) My ISP just shrug their shoulders... How can I test if the MTU path discovery problem still persists?

I'll set the MTUs & let you know here on the forum?

Thanks
Ekkas

[ekkas2]

I set the MTUs with DrTCP and this time our server ran for 3 days (a record) without freezing, however, this time the Internet stopped working on the workstations but was working on the server. After a delete history files and reboot all is fine again.

What else can I look at?

Thank you
Ekkas

[ekkas2]

How can I test to see if MTU path discovery is still broken by my ISP?
Thanks
Ekkas

[ekkas2]

After setting my MTUs in Wingate, OS & router, the freezing is definately less but it still happens. Also now sometimes a website does not open properly or times out completely, after a refresh it opens ok. On the server the pages always open properly...

Isn't it possible for the QBik guys to do something about this? This problem was introduced (I think) with version 6.1.
I know it is because of my ISP's MTU & path-discovery nonsense, but for WG to just freeze up is not nice at all.

What can I try to stop this? Timeout or other advanced settings? What is best network buffers setting for 100 users? (Now on 4096). Etc. etc., anything would help.

[adrien]

Hi

Problem with MTU issues like that, is that packets that are too big are silently dropped by your ISP.

WinGate relies on the OS TCP/IP stack for all network communications (except for NAT). For all outbound connectivity from the WinGate machine, you're also using the OS TCP/IP stack.

the TCP/IP stack really doesn't like having random packets dropped on it silently.

I sent you an email a while back about how to test MTU using ping - do a search for ping with -f and -l flags. It's possible to establish the correct MTU using a bunch of pings to narrow down to the true value.

Regards

Adrien

[ekkas2]

Thanks for the reply.
I've checked the MTU with ping and TCP-optimizer, they both give me 1454 which I've set my outgoing loadbalancing router to, WG network override & OS with DrTCP but it's still the same. In fact, I am using 3 ADSL lines, two of them (satellite routed) are MTU 1500 and 1 (fibre routed) is 1454 so I've set my whole system to the lowest denominator, which is 1454. I've requested my fibre line to be converted to satellite which would then also have a MTU of 1500 but that can happen only in a month, in the meantime my clients are a bit frustrated...

Ekkas

[ekkas2]

I've just found out one of my lines are unstable. Pings & times out every now and again, that wcould be the cause for pages timing out. Now if the freezing problem could be sorted out...

Please help me understand, even if packets are silently dropped, should/wouldn't there be a timeout of some sort to handle such inconsistencies? I just do not understand why WG would freeze, regardless of the state of our ISP or network outside of WG.

Thanks
Ekkas