На чем он? Возможно можно портировать если он на libpcap

На Си, он там отдельно модулем выносится, насколько я понял

883202 – Ethernet ring/automatic protection switch being active up on restart, fedora 17
https://bugzilla.redhat.com/show_bug.cgi?id=883202

Надо будет тогда код 801q/бриджа переписывать, да и какую версию 1 или 2 писать )

Все ж имхо это можно и на железке рядом сделать...

Hi @dimka_88  @sergey239  My problem whas that I'm using OVS (vswitch) in Proxmox, and by default only allow 1 tag. Need to confing to allow 2 tags. But I have changed to Linux Bridge interfaces. For Now all ok.
My scenario is a public net routed to this accel-PPP, and vlan per customer/client + dhcp. I have some structural doubts:

-Can I have some customers with nat connection and others with routed public ip, at same accel-ppp instance?

-What about peer to peer connections,  I haven't got ping with Customer 1 and 2, maybe the problem is that I'm sending /24 mask via DHCP, maybe /32. I understand that customers between their can play games or make calls with direct ip access?

- Any criteria about dhcp leases time? Maybe 600 seconds is low for +1000 clients?

-At the moment that I have 2 accel-ppp instances for load-balancing, I need to play wiht dynamic routing with OSPF, exporting active sessions /32?

Many thanks :)

Hi
>Can I have some customers with nat connection and others with routed public ip, at same accel-ppp instance?

Yes, you can. Accel-ppp doesn't do NAT operation so you have to configure iptables/nftables for nat subscribers

>-What about peer to peer connections,  I haven't got ping with Customer 1 and 2, maybe the problem is that I'm sending /24 mask via DHCP, maybe /32. I understand that customers between their can play games or make calls with direct ip access?

check proxy_arp configuration. sysctl -a | grep proxy_arp . also I'd recommend you to analyze traffic via tcdpump . if this problem is related to proxy_arp you will see that your server is not replying to arp requests

-At the moment that I have 2 accel-ppp instances for load-balancing, I need to play wiht dynamic routing with OSPF, exporting active sessions /32?

yes. usually ospf/ibgp is used for this scenario. you need to advertise each subscriber's /32 prefix to "upstream"/"core" routers or route-reflector (depends on your network design)

got it! not proxying arp 👍

regarding to dhcp lease time. 600 sec for 1k subscriber looks ok. you can start from this number and increase/decrease in case of issues

@sergey239 ok to all! very thankful!

BTW, you need to increase arp_cache size in linux kernel for >1K subscribers. check it here https://accel-ppp.readthedocs.io/_/downloads/en/latest/pdf/ , section 4.3

Ack!

NAK!

колеги , а чтобы при ipoe выдать по радиус статичесий Ip 10.0.128.100 что радиус должен вернуть, при pppoe я передаю  Framed-IP-Address ,  
в секцию ipoe я добавил: gw-ip-address=10.0.128.1/19

этого же достаточно ? при условии что DNS прописал а конфиге Accel.  или может кто еще что выдает клиентам ? послезное ?

Должно хватать

Достаточно. Но можно ещё и dns слать если есть желание

Если не был получен атрибут, то выдаст из конфига