Port Errors

[sirhc]

Well let me warn you in the past 3 years I have had to return (2) airFIBER units that had bad Ethernet ports.

The first unit failure led us on a wild goose chase that lasted a couple months so after that we learned to check for that.

Considering we have 20 AF24 radios that is still over 10% failure rate but you would still have to pry those 20 AF24 radios from my cold dead hands!!!!

The first bad AF24 talked fine to our laptop NIC and we pulled/pushed 700Mb if we plugged direct to the POE brick but it did not talk to the router (this was before WISP switches) which was a Cisco 2951 and the router did not show any interface errors.

When we logged into the AF via SSH and looked as there is a command that shows errors on the AF NIC.

For the life of me I can not remember the command so I sent a link to this post to TOM my CTO who will post that command up here tomorrow.

Happy hunting!

[tossman101]

The command is emacstat.

[TheHox]

I was sitting here thinking about this again. and I looked at all 4 of my AF5s.

All of them list the DATA Cable Length, except the one throwing the port errors.
It just says

< 20M

So, I would assume since the AF is unable to determine its wire lengh, it also must be detecting something wrong in the wire. Which is good news, as that might finally confirm the problem. The WISP Switch wire diag shows the wire good still, but at least this is a start.

Hopefully find time this weekend to swap ends again, and maybe even a new wire run.

[TheHox]

I am getting high latency during my peak times. During the day I can get 8ms to 8.8.8.8, now I am getting 90ms.

My provider first thought the AF5 link between him and myself was maxing out. I observed it now doing 80mbps.
During this time my latency is between 80ms-100ms to 8.8.8.8, google.com, akamai.com etc.
I then was able to login to that AF5 link, and via the GUI I can ping the slave from the master side and the average ping out of 50 pings was 1.8ms.

So, the link itself is not being maxed, but, I still do show RX DROPS counter rising, on my Netonix. Would those RX drops cause higher latency during peak usage?

My provider is not being helpful at all. this happens every night, he acknowledges it, says he will look into it, but nothing happens, and it is only getting worse as we add more subs. So I am stuck here trying to figure this out, while the problem is likely not even on my network. Basciily anytime the link pushs 40mbps or more, the latency climbs into the 80ms range, if not about 100ms. Other times, it is almost rock solid under 10ms.

[sirhc]

No, Rx Drop would not be your cause.

My guess is it is your provider since this is happening at PRIME TIME, our switches do not care if it is prime time or 2AM so if your net is fast at 2AM then slow at prime time I would say I would talk to your provider.

[TheHox]

That is pretty much what I think I have narrowed it down to.. He is on this forum, as he runs a larger wisp, but still no resolve here..

Hop 3 is my core, hop 4 is his core, but there is something in there that he is having major issues with.. Nothing else I can do besides sit here and watch.. or shop for other providers.. All of my customers are on this and dealing with this same crap right now.. /rant

[TheHox]

My provider finally cut the BS and admitted there is a ToughSwitch at the tower where the master AF5 is at.


He racked a Netonix switch there the day we setup this link, but did not switch over due to business customers being on that link. I thought he did it that night, Would have been nice to have known that a month ago... Hopefully that will resolve the issues.

netonix.png (1.6 KiB) Viewed 12080 times