Hi,
I have 7 mikrotik high sites, each with a sector antenna for clients, and a grid antenna connecting to the backbone. on each tower is an EoIP tunnel to my PPPoE server. the EoIP Tunnel is in a bridge along with the interface for the sector antenna.

on the PPPoE server (RB750GL) all the EoIP tunnels are in a bridge, the bridge is in a PPPoE server.

All bridges are RSTP

All EoIP Tunnels have unique MAC's in the range according to the wiki.

I started to notice clients disconnects and the whole network stops responding. I have swopped out routerbords on almost every high site, including the pppoe server. On closer inspection I noticed on the PPPoE server that the CPU hits 100% and the bridge interface would show 36-52Mbps traffic transmitted and all the EoIP tunnels would max out their respective bandwidth on the wireless. The whole system looks like its overloading. in that time all pppoe clients drop and no access from the high sites to the PPPoE server can be made, not even ping. this would last in the range of 10 seconds to 2 or 3 minutes at a time at completely random intervals. I thought it could be an IP conflict, can not find one. I have no clue left where to look or what could be wrong here.

when it happens again do a /tools profile to see what is causing the issue.

My guess is that it's something related to Spanning-tree..

I did a /tools profile under normal running conditions as a base line, this is what I get after watching it for a while as the values jump a bit but this is the average:

Idle : 70-90%
Management : 5-13%
Firewall : 3-8%
PPP : 1-3%
Queuing : 1-5%
EoIP : 0-1%

The rest of it is 0-1% with 19 to 21 items total.

I had a system crash this morning which resolved itself just as quickly as it started. I was able to log in remotely to the RB750GL one time and get a few seconds glimpse of the profile before I got kicked off.

Idle : 10-20%
Firewall : 40-50%
EoIP : 30-40%

During this time, the one tower on the roof of my house had 3.5-5.5 Mbps traffic on the backbone interface, the link is limited to 6Mbps. It had no internet connectivity, and no clients on it could work because all the PPPoE connections dropped because it had no access to the PPPoE server (RB750GL)

I was unable to get more info as I literally had 10 seconds to look at it before I got kicked off and could not log back in. I then made the 5 min drive to my office where the RB750GL is located, but when I got there the whole thing resolved itself.

I am also running PRTG Network Monitoring Server on my network, as I hoped it can tell me something about my issues, I set it up to poll info every 60 seconds on Pings, and wireless link strength between the links on the backbone, and on the core routers. So if any info from the network monitor can help, i will provide it if I have it.

All that PRTG is telling me at the moment is that pings increase from less than 5ms to over 1000ms to the indevidual backbone links. and then there are ping losses when the RB750GL stops responding for a few seconds. The PRTG server is in front of the RB750GL, and the backbone is behind the RB750GL.

I completely removed EoIP from my network and set it up using static routes ans normal bridges. Everything worked fine for a week and i was convinced I solved my problem, until last night.

At 6pm the network went down, same as before. I rushed out to the office and found the PPPoE server (RB750GL) at 100% CPU load, a /tools/profile showed 79% usage on firewall and the rest on management. I rebooted it, no effect, it started up and went to 100% CPU load. Then I disabled the Hotspot server running on the router, and immediately the router went idle. Put hotspot back on and CPU followed to 100%. Left it off again for 30mins and when I switched it on it was ok, CPU was fine. So I left it there.

At 9pm the network crashed again, same as before, switched hotspot off and CPU went idle. So now i removed the hotspot from that router and put it on another router.

Does anyone know what could be going on? I will specify any detail needed for diagnosis, there is so much configurations on the network, it's hard to specify it all at one time, but if you need it I can give it.

I also have an RB750GL that cpu overloads, unit becomes unresponsive and reboots from watchdog timer every hour, Im about ready to smash it with a hammer.

You are kidding? 750GL has 400MHz Mips CPU and you use it as tunnel server?
You're Fileservers are 10 year old Laptops?

Just buy a 1100AHx2 and save you a lot of time.

I figured out the issue with my unit, Im using the mikrotik web proxy to redirect to my upstream proxy, apparently 128k default max object size is too large and causes excessive flash reads/writes I reduced Mac Cache Object Size to 8k and the RB750GL unit hasnt rebooted once since. Which is just fine by me, I dont even require a cache on the mikrotik since I have a local gb connected squid x64 machine.

They should change the default max object size to 8k to avoid users running into this issue.