Nailed the issue for 95%:
All links / ethernet cables leave from CCR1009 over some 15 meter of ftp cable (all well grounded on both ends....) to a tower cabinet.
In the cabinet a 24V PoE battery fed midspan power inserter with lighting/surge protector is fit. From here some 8-10 meter of ftp (all well grounded on both ends...) run up the tower to connect to the radios. Mostly Netmetals.
The midspan PoE injector is from Cyberteam, Poland. This is a managed Gigabit version. This is actually the 3rd box that give me problems. One just died, the second had no more webserver working and became useless because of that en this last one after 2 months again fell without web server control. So my suspicion went to this device. Last night I replaced it to a brand new unmanaged gigabit net-protector and almost all fcs error disappeared from the logs.....
So this is the 3rd managed netprotector heading for the bin (with 2 unmanaged going the same before..... poor quality stuff....)
I still suspect the last remaining errors, now only on one link, are due to this new netprotector. But to be honest there are not a lot 24V multiple port passive PoE midspan injectors in gigabit version with IP management on the market. The nearest solution I can find are Netonix WISP switch but I need to convert it into 1:1 port to port switch only since the several routes have different dhcp-servers and hence these 'pipe's need to stay separated.
To reply to some of your comments;
1) If you can swap the 1100 back in and the FCS goes away, then i would agree cabling is prob not it
I am 100% its not the cabling. It also happened on almost every port, and they all have Gb connections.
2) This is true only on copper (and you can set 100M Full on each side just for testing) SFP ports can be hard coded to 1000/Full as well.
Since the midspan PoE insert and the fact that CCR sits as router in a crosspoint of links, 1:1 SFP connections are not possible. Also because we don't have the tools to make fibre cables nor can I buy prefab SFP with lengths of more than 5 meters against affordable prices....
3) In our lab we have a CCR1009 connected to MikroTik, Cisco and HP network equipment on copper and fiber without issue....unfortunately you probably got a bad CCR.
There was a batch of bad CCR1009s early on and it's possible you got one that's been on the shelf for a while as all the new CCR1009s shipping seem to be pretty solid.
I agree, I am now more looking at another midspan PoE solution.
Show your appreciation of this post by giving me Karma! Thanks.
Rudy R. Puister
WISP operator based on MT routerboard & ROS.