I am running PPPoE access concentrator. It provides connection to 1,5k customers. In recent time, i have problem with disconnects. I was suggested by mikrotik support to make new ccr for pppoe only /no NAT, no connection tracking, no firewall/. Now i am running it on ccr1036, peak CPU is 15%. Once per day i need to reboot it, because i loosing gradualy all /ppp active. After reboot i have many tunnels ok, but some has no ip address. maybe 5% of all connections has no dynamic IP in /ip address. This ccr has 100 VLANs, so 100 pppoe servers are running there. Is there any idea, hint or suggestion?
Interested thing is, that after reboot i can see some tunnels in interfaces above VLANs. I mean some tunel connections without IP address are above VLAN hierarchy
As i mention, problem occures when only 1 ccr1036 is doing pppoe-server. It is doing only pppoe-server within 100vlans /using local secrets, No firewall, no NAT, no connection tracking. I cannot run 1,5k pppoe-clients on one ccr. It is not matter of CPU, problem is when i passs 1,3k clients running. It is unable to create dynamic IP ADDRESS entry. I saw problem on this ccr1036 yesterday, when connecting 1,2k users, load is less then 2%. When less then 1,4k load is 15%. When pppoe clients is getting to 1,5k, i have CPU 15-99%.
When i run 2 pppoe server, it is OK. But it is not solution for me:
did you find any solution? We just hit the same problem today. 14 days uptime on CCR2004-1G-12S+2XS, ~1100 PPPoE sessions, simple firewall, simple queues and no problem whatsoever. Today I just moved all PPPoE session to different concentrator. After moving them back to CCR2004-1G-12S+2XS, about 50 sessions had no route in routing table (no ip address either). PPPoE sessions looked ok. Manually clearing them did not help - new sessions were established but still with no routing. Did not rebooted CCR2004-1G-12S+2XS yet.
