I Have a few questions that anyone could elaborate on and tell me if im wrong.
I have gone through a few rb’s after getting info from tracerts and picked up that there are some rb’s where the costing of the backbone links has changed.
Now my understanding is that on the wug OSPF costing for bb links are 10 and ether links between rb’s are 1.
With the changes of costing I have seen that tracing from point A to B gives 5 hops but tracing from point B to A I get 13 hops is this because of the costing ?
If it is so what is the reason for this and should this be alowed
We are getting lag spikes and time outs in games because of this and a lot of irregular pings to services.
I don’t mind if this is done for a reason but i cant see the logic in this unless its to disable a link but in that case i would say disable the link and not change the costing.
Please any info on this regard would be appreciated.
The only knowledge I know for changing costings is to increase those performing badly in order to still keep it up and transmitting some traffic at least.
I don’t see why costings would be lowered unless someone has done the same on the other side or with sufficient planning.
Anyways changes like this should be tested and published I would say.
First thing I would suggest is to scan the logs to see if you can find out who changed the costings and find out from them the reasoning for it.
Are these private ospf nodes or ctwug nodes as well and which rb’s specifically. From there we can see where to go next… perhaps test to see if the costings is causing the lag ( so have someone with lag constantly ping and then changing the costings to see if there is an improvement ). If no improvement then the revert wont stick and visa versa.
Its a good idea to manipulate the cost on router boards if we do it for the right reasons, as said Toady making the cost higher on rbs that are struggling could help some packets choosing another route taking some stress of the rb. Making the cost 5 on dual-pole links should be standard by now seeing that in most cases these links should be strong links that are under worked and in need of more traffic.
Ye but if the costing isnt done on both sides you are only benefiting the traffic in one direction… and it seems the rb’s graphire has found has only been changed one side.
Yes the costing was only done one way and only benifited in one direction so thats whats messing around with the routing between points and thats why i asked the question. I do understand that changing the costing for faster links is what we should do. But then the costing should be changed right across the board and not just on one or 2 rb’s and on both sides of the links.
For most parts, changing the costing on CTWUG links isn’t that simple, the network is too complicated. If you want to change costing, then you should ideally be able to track & model link speeds and active usage for all OSPF links, which is pretty much impossible on CTWUG.
Only solutions I can really come up with is
1: Only “OSPF Admins”, with knowledge of all interlinks between areas, should change costings, and this should be documented.
2: A standard costing used throughout CTWUG (i.e. ether=1, wlan=10)
3: Split the areas up with BGP. This allows each area to worry about it’s own internal OSPF costing, and also allows the areas to decide how they want to route to other areas (to a degree).
Option 3 is going to be a mission to get going though, but this is were I see CTWUG moving to eventually.
Call me nuts but I would totally dig option 3!!! I see this might be the future of our network (if not MPLS) , this would require some very good planning and clever guys to manage this!
The biggest issues with option 3 is going to be defining exact area borders, and getting area IP’s inline (like KFN did recently with 172.26.x.x). IIRC there were issues a year or two ago re area borders?