Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Sep 14 11:24
    pettai closed #912
  • Sep 14 11:24
    pettai commented #912
  • Sep 12 10:42
    thomas-mangin commented #915
  • Sep 12 09:15
    omkartikare closed #915
  • Sep 12 09:15
    omkartikare edited #915
  • Sep 12 07:53
    omkartikare edited #915
  • Sep 12 07:52
    omkartikare opened #915
  • Sep 11 15:30
    mattoddy commented #913
  • Sep 09 11:27
    thomas-mangin commented #912
  • Sep 08 08:12
    thomas-mangin commented #912
  • Sep 05 11:22
    pettai commented #912
  • Sep 05 10:37
    thomas-mangin commented #912
  • Sep 05 10:31
    thomas-mangin commented #912
  • Sep 05 10:07
    pettai commented #912
  • Sep 05 10:06
    pettai commented #912
  • Sep 05 10:06
    pettai commented #912
  • Sep 04 19:31
    pettai commented #912
  • Sep 03 19:42
    thomas-mangin commented #912
  • Sep 03 19:22
    pettai commented #912
  • Sep 02 16:25
    thomas-mangin closed #913
Ben Agricola
@benagricola
the change to the regex is simple i think
it's looking for some trailing spaces after the 'loopback IP label' which appear to not exist in the output from ip addr show on our machines :)
(\s+.), I'm not sure why it's not just . because the trailing info doesn't matter (even if it does exist) once the label is parsed out
also just having a quick look at the daemonisation order - exabgp runs daemon.daemonize() prior to process.start() so that explains why it's necessary to run exabgp as root in order to have the healthchecker also run as root!
Thomas Mangin
@thomas-mangin
ok - I may change this - can you please open an issue raising it. It MAY not be possible for other issues … but I can not recall and should investigate
Ben Agricola
@benagricola
yeah will do
i'll submit a PR for the regex change in healthcheck.py too
Thomas Mangin
@thomas-mangin
Thanks
Thomas Mangin
@thomas-mangin
@vincentbernat - it seems the code need to not failover if the IP is already there and perhaps intercept SIGTERM to perform the IP removal ?
Ben Agricola
@benagricola
@thomas-mangin when it correctly detects the IP already exists it doesn't error which is fine
but intercepting sigterm to remove it is probably a good idea anyway
Ben Agricola
@benagricola
heh, so I had a look at removing the IP address on sigterm as well
it's only possible if the healthcheck script runs as root
obviously if you have it drop privileges it's no longer able to remove the IP address on sigterm :D
Ben Agricola
@benagricola
@thomas-mangin is it possible to set a default med for each neighbour or template? or would that only be possible by modifying the announcement made to exabgp?
Thomas Mangin
@thomas-mangin
No .. sorry, feature request but I do not know yet how I would implement it.
Ben Agricola
@benagricola
@thomas-mangin no worries :)
Bo-Han Liao
@hankofficer
Hello everyone, I'd like to ask a question. Hope I didn't cause any inconvenience here. .u.
I tried to run to exabgp on Windows but got no luck, it seems I need to find several Unix modules in order to get it work. Do anyone have idea or tips how to make it run on Windows?
Btw I'm glad I found this software on github, this could save a lot of time. keep up the good work! :smile:
Thomas Mangin
@thomas-mangin
Sorry ExaBGP is unix only ...
To run it on windows: install virtualbox, run linux on it :wink:
Bo-Han Liao
@hankofficer
Aww okay. Thanks for help. :)
Vincent Bernat
@vincentbernat
@thomas-mangin I have some difficulties to package the new ExaBGP in Debian. qa/bin/conversation is timeoutting for all tests, except E and K
Is there some easy way to make those tests more verbose?
Or from the top of your head, is there anything that changed significantly since 3.4.13 on this part?
Thomas Mangin
@thomas-mangin
Hi @vincentbernat - I am not sure myself. The test are successful on my laptop but fail on travis-CI but pass on codeship !
I added one test and I am forking two processes per test. Could it be that I am hitting some limit somewhere ?
Vincent Bernat
@vincentbernat
I'll investigate a bit, I was just asking in case you had already a clue.
Thomas Mangin
@thomas-mangin
sorry - no clues as all the tests are fine individually
Vincent Bernat
@vincentbernat
@thomas-mangin: I am wondering if the problem doesn't lie in the two communicate()
if the daemon outputs too much, it will block
Vincent Bernat
@vincentbernat
also, the communicate() in dispatch is a bit late: if a process has more than 8 or 16kb to write, it will just block
Vincent Bernat
@vincentbernat
however, in my case, this doesn't seem to be the problem
but it would be easier to debug if the script was collecting output from both programs iteratively, instead of relying on communicate()
for example each output could be collected in a thread, or a loop with select()
Vincent Bernat
@vincentbernat

I am also getting this:

Sun, 20 Mar 2016 00:02:48 | ERROR    | 29879  | configuration | syntax error in section process
Sun, 20 Mar 2016 00:02:48 | ERROR    | 29879  | configuration | line 12: run ./api-nexthop-self.run ;
Sun, 20 Mar 2016 00:02:48 | ERROR    | 29879  | configuration |
Sun, 20 Mar 2016 00:02:48 | ERROR    | 29879  | configuration | can not execute directories "/etc/exabgp"

when /etc/exabgp exists.

Unfortunately, I don't have much time this week.
I could just skip those tests
I suppose it's important you run them, I don't need to run them myself
Thomas Mangin
@thomas-mangin
Thank you - I will look into this /etc/exabgp issue
You can indeed skip the tests - I run them on my laptop before release
They do not need to be run on the build
Thomas Mangin
@thomas-mangin
@vincentbernat /etc issue resolved on master - silly pebkac. Exa-Networks/exabgp@2eda799
Thomas Mangin
@thomas-mangin
Screen Shot 2016-03-20 at 09.34.02.png
Go understand !
sigh !
Thomas Mangin
@thomas-mangin
@vincentbernat search in /etc/exabgp backported too
Thomas Mangin
@thomas-mangin
@vincentbernat my laptop and codeship have no issue with running the tests
If it is failing on your machine, could you just find one of the failing test (not green) when your run ./qa/bin/conversation
and then use two terminals