Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Aug 14 17:35

    thomas-mangin on 4.2

    bug by rstcheck (looks like it)… (compare)

  • Aug 14 17:10

    thomas-mangin on 4.2

    add zipapp support updating version to 4.2.11 (compare)

  • Aug 14 17:10

    thomas-mangin on 4.2.11

    add zipapp support updating version to 4.2.11 (compare)

  • Jul 24 10:13
    thomas-mangin locked #988
  • Jul 24 10:12
    thomas-mangin closed #988
  • Jul 24 10:12
    thomas-mangin commented #988
  • Jul 23 16:54
    bopopescu opened #988
  • Jul 20 07:58
    thomas-mangin commented #987
  • Jul 20 07:57

    thomas-mangin on master

    Fix crash when daemonizing and … Merge pull request #987 from vi… (compare)

  • Jul 20 07:57
    thomas-mangin closed #987
  • Jul 19 18:56
    vincentbernat opened #987
  • Jul 08 14:32
    alexejli commented #668
  • Jul 06 12:37

    thomas-mangin on master

    update links (compare)

  • Jun 24 11:06
    N-Mi commented on 4989496
  • Jun 19 16:26
    thomas-mangin closed #986
  • Jun 19 16:26
    thomas-mangin commented #986
  • Jun 19 06:44
    reza-ghazvini commented #986
  • Jun 19 06:42
    reza-ghazvini commented #986
  • Jun 18 14:37
    reza-ghazvini commented #986
  • Jun 18 14:37
    reza-ghazvini commented #986
Thomas Mangin
@thomas-mangin
There is never a good time to stop to be a student :-)
Nicholas Hart
@hdb3
TTFN - thanks for the publicity
Thomas Mangin
@thomas-mangin
You are welcome. Enjoy the weather if you can !
Nicholas Hart
@hdb3
i mesnt to ask - why did you use the next-hop as the key for prefixes in the json? - i would have thought it should be a value, not a key.....
Thomas Mangin
@thomas-mangin
If you look at how routers are set-up. Every EBGP router will change the next-hop to be within the ASN (to garantee the next-hop is in ISIS/OSPF and can always be reached). So while updates the attribute will change a lot, the next-hop will not. It is an optimisation.
Nicholas Hart
@hdb3
also - if you don't do prefix packing then the MRT table breaks the relay script - i thought it was a performance bug in your code for very dense routes - but the problem went away when i switched to JSON and the packed syntax. It might still be worth looking at though.... i think that ii was the text format output which is broken but i couldn't see an obvious casue and i don't really do python. (any more....)
Thomas Mangin
@thomas-mangin
The text format is really deprecated but I am not killing it to not break people's installation which rely on it for simple things.
Nicholas Hart
@hdb3
Ahhh - BTW I don't agree entirely - you can redistribute the external links in IGPs - which makes somethings easier - and that is without getting into MPLS....
I prefer to do the latter and I know some ISPs do too, or something equivalent.
Thomas Mangin
@thomas-mangin
If you distribute BGP into your IGP you are looking for trouble as it will massively impact your network convergence time.
Nicholas Hart
@hdb3
NO - not full table - just the connected exyternals
not from BGP at all.
Thomas Mangin
@thomas-mangin
You can indeed add the /32 in ISIS ..
but really you can not be sure if the route you go has the /32 as the next-hop - it could be any IP on the router
Nicholas Hart
@hdb3
Most of my testing is with vMX and VIOS....
Thomas Mangin
@thomas-mangin
You mean VyOS :-)
Nicholas Hart
@hdb3
well - you do need to configure your ASBRs properly ! ;-)
Thomas Mangin
@thomas-mangin
or Cisco VIOS :-)
Nicholas Hart
@hdb3
Cisco
IOS works in KVM
Thomas Mangin
@thomas-mangin
Our core is Arista ATM
Nicholas Hart
@hdb3
IOS-XR is more problematic - seems only to be ok in VMware - i couldn't get it working in libvirt.
Thomas Mangin
@thomas-mangin
must be the drivers for the NIC :-)
Or am I wrong ?
Nicholas Hart
@hdb3
yes - but not in an obvious way....
I think that they are messing with the PCI inititialissation
DO you do any transit?
I have some links with TNP and B4RN but they are just stub ASes...
Thomas Mangin
@thomas-mangin
Friends and Family only :-)
It depends if they want full routing, etc.
Nicholas Hart
@hdb3
how many family members with their own ASN?
Thomas Mangin
@thomas-mangin
Many friends :-D
Nicholas Hart
@hdb3
cool - i really must depart - at some time later i may have more test observations - i think that exaBGP is probably faster than GoBGP in some cases!
Thomas Mangin
@thomas-mangin
Doubt it !
If so, I should be able to fix goBGP in a few hours :-D
All my recent code is in Go and performance are good !
perhaps with pypi on single core ExaBGP will hold its own !
Nicholas Hart
@hdb3
No - really GoBGP is quite broken.
I could fix it too, but I'm not going to.
My future is probably in Rust.
Thomas Mangin
@thomas-mangin
I like what they want to do with Rust but I will continue to use Go
Nicholas Hart
@hdb3
BTW Haskell is truly beautiful for network programming - ADDPATH was about 100 lines extra, including reasonable semantics. zserv API took a week (quagga FIB interface).
Only problem is the garbage collection, and laziness is hard to contain without care. But for proof-of-concept and prototype s nothing else gets close. As long as you already know Haskell. ;-)
Thomas Mangin
@thomas-mangin
GC is good with Go now.. Only started to use it at 1.7 when the problem was under control / good enough. Fast cryto is a must for me and again Go is good at that (Rust too).
with many algorithm optimised in ASM
Nicholas Hart
@hdb3
I'll send you the GoBGP results when i retest it shortly...
Thomas Mangin
@thomas-mangin
Cool thanks
Nicholas Hart
@hdb3
BTW I tested GoBGP about a year ago - it may have improved - i would not go public on results without careful checking....
pizu
@pizu
hello
the above are 2 links with the errors am having