We have been having this type of problems for a long time, we have tried all RouterOS versions from 6.32 to 6.42.3 through all the intermediate ones "BUGFIX". It is always the same problem since we started using MPLS. I ask you to help me because it is a very serious problem that is causing me many headaches.
I proceed to explain and expose the case:
We use Mikrotik as a provider, with OSPF, BGP, MPLS, VPLS and PPPoE.
In this case we have this scenario.
This last problem has occurred in the router "SERVER PPPoE01", although last week it happened in the "SERVER PPPoE02".
The router "SERVER PPPoE01" has two ethernet connected against the "CORE_MPLS05" in a bonding mode balance-rr and over the bonding a VLAN with a /30 by which the OSPF, LDP, MPLS, VPLS protocols are raised.
When the problem occurs in the router "CORE_MPLS05" it is appreciated that:
- You can not see the router "SERVER PPPoE01" in neighbors.
- The OSPF is down.
- The LDP is down.
- The bonding interface, the VLAN and the 2 ethernet do not have any reception packet RX instead if it shows TX output packets.
- Does not ping to the other / 30 router.
- CPU and memory are normal.
When the problem occurs we have to connect to "SERVER PPPoE01" directly to a free ethernet to do some troubleshooting and from within the "SERVER PPPoE01" router this is what we see:
- You can see the "CORE_MPLS05" router in neighbors.
- OSPF is UP but without any loaded route.
- The LDP is UP but in "LDP neighbors" in the neighbor of the router "CORE_MPLS05" there are no addresses and "Forwarding table", "Local Bindings", "Remote binding" are empty.
- The interefaz bonding, the VLAN and the two ethernet have RX reception packets but only the VLAN has for TX output packets and both the bonding and either of the two ethernet do not show any TX output packet.
- It also does not ping the other / 30 router.
- And the VPLS that is raised against the third router is obviously down.
- CPU and memory are normal.
After observing and diagnosing the problem, we made a supout.rif that we attach with the name "supout.rif".
As it is not the first time it happens we know how to fix it temporarily, which is doing a shut no shut in the ethernet interface. But in a couple of days or a week it happens again and we have to repeat the shut no shut of the interfaces. The only way that it does not happen again is not using MPLS and LDP but in this and other cases it is not an option. We have this same scenario in several sites with the same configurations, but it only happens in a small percentage of routers.
Then after doing this first supout.rif we proceed to make a shut no shut in the ether1 of the router "SERVER PPPoE01" with this I raise everything for that interface (OSPF, MPLS, VPLS) but losing half of the packets, since the ether2 continued hanging then we took out the second supout.rif with the name "supout-after-eth1downup.rif"
The next thing was to shut down the ether2 of the "SERVER PPPoE01" router with everything running and fixed, until the next time it happens again that it will probably not take more than one or two weeks. We also did a supout.rif with the name "supout-after-eth2downup.rif"
*** Doing shut no shut in the interfaces of the other router "CORE_MPLS05" does not solve the problem, the interfaces of the "SERVER PPPoE01" fall and raise but the problem is not solved, it is only when the shut no shut is done in the "SERVER PPPoE01" router.
Attached I leave the supout files and I hope that with this help we can find the solution.
I'm waiting for your comments or if you need more information.
Regards.