I have replaced RIP with OSPF on my small network.
After some time, network outages are more frequent.
Everything that I can blame is OSPF with these errors.
What can be wrong here?
Please post a network diagram. Any notations with additional information like IP, interface types and presence of tunnels along with the router configs any plain text will greatly help.
Also post an example of your OSPF configuration from one of the routers…
/routing ospf export compact
Dude screenshots will be best diagram 
I´m sorry but names are not in English. You will reorganize MT routers by CPU/memory status on dude maps.
Searching forum I found that there was people that had same or similar problems like me, but nobody found a solution. One idea is to use NBMA with neighbours list and avoid using broadcast. Lucky for me, OSPF make these problems every day or two, not every 5 minutes
Problem is always in backbone area, others are fine (nothing in log).
Just as a starter, you did redistribute connected a lot. It looks like that is how you’re trying to bring in your local networks and interfaces. It’s far better to define the necessary networks and interfaces. You can set the interface to passive to prevent unwanted neighbor-ships.
I’ll have to spend some additional time to digest the rest of the drawing and configs in the mean-time are you able to elaborate on your usage of redistribute connected over defining the interfaces and networks?
On every MT there is bridge with ip 10.0.x.129/25. Bridge ports are eoip and AP interface.
This IP class is used for cpe management.
Without redistributing connected routes I don’t know other way to make this class available to rest of network.
routing ospf interface add disabled=no passive=yes interface=MyFirstEoIPInterface network-type=broadcast
routing ospf network add disabled=no network=10.0.x.29/25 area=backbone
Alternatively, your routers should use a loopback interface with a /32 for management unless there is a true need to bridge. I’d still implement a /32 even if I needed to bridge traffic to the device. The advantage of this approach is the /32 is distributed by OSPF and if a device has multiple paths in the network they will all be known. When a link fails you will still be able to manage the device.
Think of 3 routers connected like a triangle. If you make all links routed links and use a /30 for each you’d have a nice little network. You can of course ping or manage the device from any of the IPs you’ve assigned but if a single link goes down those IPs become unreachable (likely). If you add a /32 to each device via a bridge with no bridge ports and just an IP to it and it to the OSPF instance it will propagate through the network as reachable via all links. You can then manage that device by that IP. I typically save a /24 or larger if needed in a network design for these “loopback” IP addresses. I carve that subnet up into individual IP addresses or /32’s that I assign to each layer 3 device.
Redistributing connected without routing filters can get you in trouble. Typically this is seen with tunnel technologies, EoIP would be one. This is commonly seen because a specific route to the tunnel destination is learned over the tunnel itself. You probably aren’t seeing this as the tunnel would constantly flap and I think you are using EoIP locally within the network. Either way, I’d use redistribution only when absolutely necessary.
I’m afraid that we don’t understand each other. Do you suggest me to use bridges and some_ip/32 classes as OSPF router ID?
A best practice: use loopbacks for managing a router. Yes the /32 could be used as a router ID. More inportantly I’d recommend a design change that gave each router a /32 on a bridge interface with no bridge ports. In Cisco we’d call this a loopback. You then add it as a network and interface to an OSPF area instead of redistributing connected.
To focus on your problem more specifically. The routers with those errors are seeing traffic from hosts on the 192.168.0.0/24 network and they don’t have an active neighbor relationship. It’s hard to tell from your configs and diagram but is that network bridged throughout your network? If so a link may be dropping some of the OSPF multicast packets along the way intermittently. I have seen other people “fix” OSPF instability by using NBMA over wireless links and statically defining neighbors. I think a better long term solution is to use the routers you have and build more point to point routed connections and less bridged, stretched, networks. Use WDS or leverage multicast-helper on wireless links as needed over changing how OSPF works.
If a network needs to be included in OSPF don’t use redistribute connected unless you absolutely have to. Define it as a network and assign it to an area.
I didn’t read too deep into the information, but I did notice that you have several OSPF areas defined.
Your network is pretty small as far as requiring OSPF areas goes.
If you don’t have a very firm grasp on what areas are and what they do, then I strongly recommend using a flat backbone area while you learn OSPF.
The fact that you had to ask how to originate interfaces into OSPF natively instead of redistributing connected routes tells me that you probably don’t know all the in’s and out’s of the various behaviors of OSPF areas. This is not an insult, but an observation.
Yesterday I ‘fixed’ ospf by disabling all router to be elected as designated router, creating loopback interfaces for router id and setting all interfaces to ptmp mode. Now ospf work more stable.
There is one new problem. One route 192.168.3.0/24 is not redistributed. Redistributed static and connected routes is turned on. How to ‘fix’ that?
^^ Great, as long as the segment isn’t a broadcast one in OSPF and neighbors show the current state for the segment type then you’re good to go. Excellent work on the loopbacks. Make sure they are advertised in OSPF.
Just to make sure you’re tracking us w/loopbacks here is a video that shows what we are talking about. This is of course about Cisco so as you’ve probably already learned we use a bridge interface without any bridge ports instead of the special loopback interface that is available to Cisco devices. https://youtu.be/pZho6OQYYxI
^^ Did you mean to say the redistribution is turned on? Start by disabling those features. Define each network you need advertised in OSPF with commands like this:
/routing ospf network add network=192.168.3.0/24 area=backbone
Yaknow, one would think that Mikrotik would implement dummy interfaces in ROS since that’s been in Linux since forever… It sure would be nice to have an actual loopback type interface instead of this work-around.
10.255.255.x is used as router ID and everything is ok about that as I can see. I can ping routers by loopback IP as cisco guy do.
Did you mean to say the redistribution is turned on?
Yes