Winbox support much appreciated and bench test so far so good, except the following bugs:
Upgrading from 2.9.18 loses all BGP instances except for the default instance
/routing filter set-prepend has now changed to set-bgp-prepend. Winbox GUI set-bgp-prepend is broken (you can add it in winbox, but you cant see it in command line and vice-versa) I presume whatever it says in the command line would be the accurate one so we will avoid filters in the GUI at this stage
Now the good news:
Redistributing BGP to OSPF is now fixed. (but this was not mentioned in the change logs?)
An upgrade for 2.9.18 to 2.9.19 to 2.9.20 did not lose any instances on our development box … maybe 2.8.18 to .20 does this only …
Changelog does not show package-test changes. Probably because its still under development and too many changes being done at the same time.
Our MAJOR bug with MT to MT bgp has been fixed in .20! Previously routes were injecting/dejecting in a loop so fast it was causing 100% cpu forever. I am glad to see this fixed and we can finally evaluate for production!
Good work MT team - thank you for your hard work as well.
We also cannot get any BGP ‘out’ filtering to work. This includes BGP prepending etc. Have tried applying the BGP ‘out’ filter to both the peer and/or the instance and it appears nothing takes effect… or has ‘the way you need to do it’ changed ?
I seem to have no troubles with the BGP out filtering in my lab setup. Have you tested in both a lab setup and production and getting the same results?
No, not production as yet, only in the lab. Using a fresh install of 2.9.20
Ive tried the same filtering setup from 2.9.18 that used to work. I am guessing it may have changed slightly… as it was a bit buggy. (Eg in 2.9.18, ‘out filtering’ only worked when applied at the ‘instance’, it did not work when applied at the ‘peer’)
I have tried ‘out filtering’ at both the ‘peer’ and the ‘instance’ in 2.9.20. It advertises OK but it seems like it will not ‘filter out’ static routes in the RIB that we do not want to advertise (eg null routing of 10.0.0.0/8 gets advertised to BGP peer with a filter in place trying to stop 10.0.0.0/8 )
Could you please post an example of how to stop advertising a static route to BGP so I can see where were are going wrong ?
Using the invert and discard, I am only getting that block to announce now. The setup I have done to test this, I set up other blocks within that range, and more blocks outside of that range (10.0.0.0/8) and only the public addresses are getting announced.