#calico

/

      • eepyaich_ joined the channel
      • eepyaich has quit
      • Sail0r joined the channel
      • piavlo has quit
      • diga joined the channel
      • piavlo joined the channel
      • piavlo has quit
      • lukasa_away is now known as lukasa
      • Sail0r has quit
      • piavlo joined the channel
      • beddari
        pic of our current simple dev setup https://www.lucidchart.com/publicSegments/view/...
      • its not literaly only 4 boxes hehe as that would be too small to expose everything we test, but this is for others to understand the simplicity of it all ;)
      • Sail0r joined the channel
      • arunpyasi joined the channel
      • slaws has quit
      • diga has quit
      • eepyaich_ has quit
      • diga joined the channel
      • qwebirc63719 joined the channel
      • qwebirc63719 is now known as anton
      • anton is now known as biwwy
      • biwwy is now known as biwwy0
      • biwwy0
        hi guys!
      • matt_dupre
        hi
      • biwwy0
        how to properly install calico_kubernetes plugin on coreos? as i see from source k8s tries to look in /usr/libexec/kubernetes/kubelet-plugins/net/exec/, it is constant for k8s and i don't see how to change this path. and in coreos /usr is readonly squashfs volume
      • so can't really put binaries there
      • --network_plugin_dir is not supported in kubelet 1.0.6
      • matt_dupre
        biwwy0: Are you following the Kubernetes on CoreOS guide (https://github.com/kubernetes/kubernetes/blob/m...
      • biwwy0
        matt_dupre: i've read this one of course. it implies that calico_kubernetes is located in /etc/kubelet-plugins and that kubelet is started with --network_plugin_dir flag, which doesn't exist in kubelet anymore i guess
      • NetworkPluginDir is a constant in kubelet source code afaik and it's not implemented yet to be changed
      • via flag
      • as in this guide
      • matt_dupre
        Most of our Kubernetes expertise is on PDT, so they won't be online until later
      • slackbot7
        <shaun.crampton> The guide says it applies to the code from master, maybe you need truly bleeding-edge k8s?
      • matt_dupre
        Or perhaps something slightly less recent - maybe the commit where that doc was last updated
      • biwwy0
        slackbot7: yep, that's my plan to try now
      • matt_dupre: will try browsing kubernetes commits and see what they have changed about it
      • still network_plugin support is in alpha stage in kubelet
      • also, i have another question. is it possible to run calico node and use authentication against etcd datastore?
      • python-etcd does support it, all needed to be done is to add few lines of code to calico_containers/pycalico/datastore.py if i'm correct
      • matt_dupre
        I don't think we've done that yet - we'd be happy to take a pull request though
      • slackbot7
        <shaun.crampton> The Calico agent, Felix, also connects to etcd, that would need to be changed too
      • biwwy0
        slackbot7: oh good to know
      • i just started trying calico yesterday will see what i need for my setup
      • oh yeah, https://github.com/kubernetes/kubernetes/pull/1... that's my issue. will need to build kubelet myself
      • slackbot7
        <shaun.crampton> It might be easier to get it running on another OS to start with
      • beddari
        biwwy0: willing to help test any PRs for adding auth, we've been thinking the same
      • biwwy0
        beddari: i've modified datastore.py for my use, testing it now. it fails build tests for now. i need to grasp how tests are conducted
      • beddari: didn't have time yet
      • beddari
        np
      • say when :D
      • biwwy0
        sure :D
      • lukasa is now known as lukasa_away
      • diga has quit
      • diga joined the channel
      • biwwy00 joined the channel
      • biwwy0 has quit
      • biwwy00 is now known as biwwy0
      • lukasa_away is now known as lukasa
      • is anybody coming here to dublin next week? for linuxcon + cloudcon + mesoscon?
      • diga has quit
      • slackbot7
        <shaun.crampton> Yes, we’ll be there!
      • <shaun.crampton> We’ll have a booth at Mesoscon and Neil’s speaking at LinuxCon
      • biwwy0
        slackbot7: nice! see you there then. guinness is on me :D
      • only the best authentic pubs to be shown :)
      • tonycoffman__ joined the channel
      • arunpyasi has quit
      • beddari
        default EtcdAddr for calico agent should perhaps not be localhost, but 127.0.0.1 / ::1
      • resolving localhost is then not in the equation :)
      • slackbot7
        <shaun.crampton> I’m not sure why we did it that way. Maybe it resolves to ::1 on an IPv6-only system?
      • <shaun.crampton> I.e. as long as your hosts file is correct, we’ll do the right thing
      • beddari
        yep you are right :)
      • I prefer not to depend on the hosts file though
      • or in other words, as explicit as possible
      • slackbot7
        <shaun.crampton> If you want to be explicit, you can set the etcd server to use in felix.cfg ;)
      • beddari
        yup I am, and why I asked ... defaults for our puppet module
      • slowed down a bit last days but we'll get there :) https://github.com/norcams/puppet-calico
      • bostondriver joined the channel
      • I had a detour writing a dependent component first, which turned out quite nice https://github.com/norcams/puppet-named_interfaces
      • not testing with v6 yet though, but the plan is a vagrant+libvirt env on RDO with 1 master, 2 compute and a cumulus vx gw
      • slackbot7
        <shaun.crampton> Cool
      • <shaun.crampton> What’s the “cumulus vx gw” for?
      • beddari
        just to show config of a cumulus+quagga BPG gw
      • and have it be the RR
      • cumulus vx is the appliance version of cumulus linux
      • slackbot7
        <shaun.crampton> ah ok, I saw vx and thought “we don’t use vxlans"
      • beddari
        "virtual eXperience" duh ;)
      • straight off the marketing dept. I'm sure
      • slackbot7
        <shaun.crampton> he
      • <shaun.crampton> *heh
      • lukasa is now known as lukasa_away
      • Marquis42 joined the channel
      • lukasa_away is now known as lukasa
      • dnorris joined the channel
      • Sail0r has quit
      • dnorris has quit
      • dnorris joined the channel
      • lukasa is now known as lukasa_away
      • lukasa_away is now known as lukasa
      • e_gordon has quit
      • lukasa is now known as lukasa_away
      • SpikeCurtis joined the channel
      • dnorris has quit
      • lukasa_away is now known as lukasa
      • dnorris joined the channel
      • dnorris has quit
      • Sail0r joined the channel
      • dnorris joined the channel
      • dnorris has quit
      • piavlo has quit
      • dnorris joined the channel
      • diga joined the channel
      • piavlo joined the channel
      • caseydavenport joined the channel
      • Sailor-2gether joined the channel
      • Sail0r has quit
      • biwwy0 has quit
      • Sailor-2gether has quit
      • lukasa is now known as lukasa_away
      • <mikezupan> anyone ever see this before with docker in kubernetes.. a container has no route table then
      • <mikezupan> ``` # ip addr 1: lo: LOOPBACK,UP,LOWER_UP mtu 65536 qdisc noqueue state UNKNOWN group default link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 127.0.0.1/8 scope host lo valid_lft forever preferred_lft forever Segmentation fault (core dumped) ```
      • <paul.tiplady> If you're using busybox, their implementation of 'ip' is broken; try ifconfig
      • <mikezupan> ok.. it has no ip
      • <mikezupan> i rebooted a node and i think calico isn’t releasing ips
      • <mikezupan> ``` 2015-10-01 18:15:03,246 5517 ERROR Cannot assign IPAddress 192.168.163.6 Traceback (most recent call last): File "string", line 285, in _assign_container_ip File "/code/build/calico_kubernetes/out00-PYZ.pyz/pycalico.ipam", line 607, in assign_ip File "/code/build/calico_kubernetes/out00-PYZ.pyz/pycalico.block", line 216, in assign AlreadyAssignedError: 192.168.163.6 is already assigned in block 192.168.163.0/24 ```
      • <paul.tiplady> Sounds like this issue: https://github.com/projectcalico/calico-kuberne...
      • <paul.tiplady> Does killing the pod (assuming there's an RC to recreate it) fix the connectivity issue?
      • <mikezupan> let me see
      • <paul.tiplady> Also, just to check -- what version of calico-kubernetes are you using?
      • <mikezupan> yep re-creating the RC fixes it
      • <paul.tiplady> That issue should be fixed in calico-kubernetes v0.2.0: https://github.com/projectcalico/calico-kuberne...
      • <mikezupan> ok i’m using whatever comes with calico/node:v0.7.0
      • <paul.tiplady> Looks like a new issue then
      • <caseydavenport> @paul.tiplady: Using v0.2.0 is not going to fix that by itself. You also need to enable Calico IPAM instead of using Docker IPAM.
      • <mikezupan> @caseydavenport: how do I do that?
      • <spikecurtis> @caseydavenport: he is using Calico IPAM --- that's our error message in the traceback
      • lukasa_away is now known as lukasa
      • <caseydavenport> Ah, that is interesting, and definitely warrants a deeper look.
      • <caseydavenport> Might be a separate issue from #34
      • <spikecurtis> @mikezupan @paul.tiplady @caseydavenport can I nudge us over to the #kubernetes channel?
      • <caseydavenport> Yep, was just going to suggest that.
      • logan2 has quit
      • logan2 joined the channel
      • tonycoffman__ has quit
      • bostondriver has quit
      • dnorris has quit
      • dnorris joined the channel
      • dnorris has quit
      • dnorris joined the channel
      • dnorris has quit