Data dropping under load, no DHCP

[opalit]

By chance you do not have the DS check boxes checked under the ports Tab for any of the ports do you? *see below*

DS.jpg

No I have not got those ticked.

[sirhc]

Not sure then.

As I said in a earlier post somewhere Eric is tied up with trying to finalize the DC switches for Monday so it will be later this week until he takes a look at DHCP Client.

For now statically assign your switch.

[sirhc]

I use these switches to run my WISP office and the RF Armor/Netonix warehouse where we use a DHCP server, all my devices get a DHCP address fine.

At my WISP I think it is an EdgeROUTER handing out DHCP and at the warehouse it is a NanoBridge in NAT mode.

I have not heard of this issues with other users getting DHCP addresses through the switches so maybe your issues is somewhere else?

[opalit]

The DHCP server is a on Mikrotik RB2011UAC RM, I have tried a RB1100Ahx2 that has been in service for 4 years and a RB750 , they are all the same with your switch, I have proved the issue is with the combination of the wisp switch and MikroTik RB DHCP.

I have proved this by bypassing the wisp switch ( device gets IP ) and if I use a Microsoft SBS server as the DHCP your switch gets an IP.

Only when communicating with a MikroTik routerboard DHCP server does there appear to be an issue.

[opalit]

P.S. there is fairly heavy traffic, 759GB in 4 days passing through the backhaul port maybe this has something to do with it.

[sirhc]

P.S. there is fairly heavy traffic, 759GB in 4 days passing through the backhaul port maybe this has something to do with it.

No, my towers pass more traffic.

There are also other people using MT routers with the switch getting a DHCP address such as jportman so I am not sure what the deal is with your setup?

I know that Eric has tested the switch DHCP client with MS servers and Linux/BSD based severs before and MT is built on a Linux or BSD branch.

I will ask Eric to confirm the switch DHCP client did not get broken since he last checked it later this week.

The amount of traffic going on the switch has nothing to do with anything related to this. I have explained this so many times I need to create a post and link to it.

The Switch Core is a stand alone product just like any other switch made by any manufacturer. There is a separate CPU that is used to run Linux to drive the web page UI and CLI that that is used to configure the Switch Core.

The Switch Core code is burned into the chip and the switch core has it's own separate processors to handle switching, we have little to no control over the Switch Core itself.

It is the little Linux CPU that drives the UI which is PRIMARILY used just to configure the Core that is failing to get a DHCP address, the question is why, is it something we or MT is doing wrong has yet to be determined.

[Eric Stern]

Someone else reported a problem with Mikrotik DHCP server, there was a change made in 1.3.0rc8 to address it. I can't recall if the original person reported whether it worked or not, but you can try 1.3.0rc11 and see if it helps.

[harold]

Someone else reported a problem with Mikrotik DHCP server, there was a change made in 1.3.0rc8 to address it. I can't recall if the original person reported whether it worked or not, but you can try 1.3.0rc11 and see if it helps.

I don't have a need to pass DHCP traffic in our network through one of these switches but I hooked up an RB2011 and RB411 to test. The switch running version 1.3.0rc11 passes DHCP traffic from the Routerboards just fine. Version 1.2.0 did not work when tested with the same Routerboards. Looks fixed to me.

[opalit]

1.2.4 is on mine, can I risk a beta version on a hot network.

I am still getting this port lockup, customers are complaining that internet is on/off every few minutes when busy, I have disabled loop protection, it is fine during the day when traffic is light and it is not always the same port, it is random.

[sirhc]

If you have disabled Loop Protection then I am not sure the issue you are experiencing is even an issue with the switch as you would think many people would be on here yelling with the same issue?

Having ports randomly drop out would be pretty hard to not notice in thousands of installations. As far as you saying heavy load a lot of my switches are passing close to 1 Gbps or over from 9AM until about 10PM and I have no issues. To be totally honest I do not think anyone including me is putting enough traffic on any of these switches to make them even break a sweat. They are line rated switch cores so if you think you hare having issues at heavy load times I would look elsewhere, just a suggestion. We tested these cores when we first put one together and we were running millions of pps across the core fabric with no hic ups.

As far as using v1.3.0rcX in production that is your call, I feel it is fine.

Your going to have to nail down exactly what is happening. I would go at this objectively and break it down into simple steps. If you come into this with a predisposition that it is the switch and it turns out not to be the switch you may miss clues as to what the real cause is.

As far as I know the know issue we are working on is SFP module error handling which under certain circumstances requires the SFP Module to need removed and then reinserted to work. This issue would not cause the behavior you are experiencing.