Reply To: Is 1.0.beta12 safe and stable for production use?

Home Page Forums Network Management ZeroShell Is 1.0.beta12 safe and stable for production use? Reply To: Is 1.0.beta12 safe and stable for production use?

#48721

jacobsa
Member

One other thing I have found with Net Balancer, that I cannot seem to resolve is external services in seem to drop packets. Not sure if you guys can test this on your services for me, but I can replicate it on a number of different Zeroshell beta12 boxes I have out there, with different services (pppoe links & static IP services).

For example:

Run a constant ping to both public IP’s of the two Net Balanced connections from another third party site, external to the Zeroshell in question.

When Netbalancer is in Load Balance & Fail Over, no matter what my weighting scenario is, I seem to lose a considerable amount of packets on the same pppoe link (4-5% over a decent time span) INBOUND only, where my other link is always fine. When I say inbound only, if I configure a rule to force traffic out via the link in question from an internal PC, I can confirm its going via that link, and never see the packet loss. Only inbound.

In all of my configurations it seemst to be a PPPOE link that always feels it, not the static IP service. I have changed weights, re-configured the PPPOE connections so they are in a different order (ppp0, ppp1, etc), but the problem remains. I can produce this on both a beta11 & 12 release I have out there right now. Even if I configure the link in question with a higher weight, it still see’s the packet loss inbound.

I have confirmed it is not just UDP packets as well, as I have tested with inbound TCP services (SMTP server, telnet, etc), and they suffer as well.

If I change the NetBalancer configuration to Failover only, the packet loss instantly goes away.

I even went as far as making a NetBalancer rule saying anything on the ppp0 link inbound use the ppp0 link outbound in an attempt to fix it, but no help. 🙂

This is a real pain when you are monitoring links for customers using a ICMP ping test. As I keep getting notified they are down, when they are not, its just this issue I have found.

I can re-produce this for any of you if you wish to see it. It is happening as I type this on the same link I am using. 🙂

If you guys could test your own links and see if you see the issue, or throw me any ideas to resolve that would be great.

Cheers,

Aaron