When provisioning using Vagrantfile and default centos/7 box w/ NUM_NODES greater than 1, the process hangs at TASK: [node | Get the node token values].

The problem appears to be the same as reported in https://bugzilla.redhat.com/show_bug.cgi?id=1242682 and https://bugzilla.redhat.com/show_bug.cgi?id=1240613.

The following change addresses the issue for me:

diff --git a/ansible/vagrant/Vagrantfile b/ansible/vagrant/Vagrantfile
index f6505a4..7d40fa5 100644
--- a/ansible/vagrant/Vagrantfile
+++ b/ansible/vagrant/Vagrantfile
@@ -128,6 +128,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
         ansible.groups = groups
         ansible.playbook = "./vagrant-ansible.yml"
         ansible.limit = "all" #otherwise the metadata wont be there for ipv4?
+        ansible.raw_ssh_args = ['-o ControlMaster=no']
       end
     end

@@ -137,6 +138,7 @@ Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
       ansible.playbook = "../cluster.yml"
       ansible.limit = "all" #otherwise the metadata wont be there for ipv4?
       ansible.tags = ansible_tags
+      ansible.raw_ssh_args = ['-o ControlMaster=no']
     end
   end
 end

I'm getting a lot of questions about what the merge bot is doing. I think if we could make that inner goroutine publish a status, that would help a lot.

@eparis any time to make this happen? :)

In particular, each release branch has it's own e2e test validating the release branch. Merges to the release branch shouldn't be blocked if e2e tests at HEAD are failing, and merges to HEAD shouldn't be blocked if tests against the release branch are failing.

We currently have 2 release branches (release-1.0 and release-1.1) that should have separate merge-bots configured.

/cc @lavalamp @eparis

under roles/nodes/tasks/packageManagerInstall it doesn't setup the Extras repo or install for RHEL7.

/cc @rrati

This is a feature request, feel free to close it if you don't think is needed or desired to have.

https://github.com/kubernetes/kubedash/blob/master/deploy/kube-config.yaml

I propose a "CLA affirmed by human" label that @googlebot won't flip, so that the submit queue can have two positive CLA signals to go off of. (It's frustrating to fight @googlebot on cherry-picks and known-good committers.)

cc @brendandburns @jlowdermilk

Vagrant Installed Version: 1.7.4

Running ansible vagrant with the openstack provider I hit the error below after doing vagrant up. A quick google suggested a workaround but not sure if thats acceptable. I'll submit a pr just in case it is.

/Users/jamesrawlings/.vagrant.d/gems/gems/vagrant-openstack-provider-0.7.0/lib/vagrant-openstack-provider/action/provision.rb:34:in `run_provisioner': Catched Error: Catched Error: uninitialized constant VagrantPlugins::Shell::Provisioner (NameError)
    from /opt/vagrant/embedded/gems/gems/vagrant-1.7.4/lib/vagrant/action/warden.rb:95:in `call'
    from /opt/vagrant/embedded/gems/gems/vagrant-1.7.4/lib/vagrant/action/warden.rb:95:in `block in finalize_action'    

Workaround came from..
ggiamarchi/vagrant-openstack-provider#240 (comment)

kubernetes/kubernetes#15452 (comment)

That PR was merge and 2 hours later merge-bot decided to retest it again.

cc @eparis @lavalamp

I'm looking to see what the effort would be to write an F5 integration. Currently my enterprise uses RHEL Atomic hosts with an F5 load balancer. We can access the F5 API to programmatically create virtual servers.

On AWS, I can specify the "LoadBalancer" service type, and it auto creates the ELB + configures. I'm looking for a way to do exactly the same thing. I am 100% cool with doing the effort, just looking for some strategy around how / where / and integration.

Is this pluggable where I could mirror how it works on AWS? Or do I need to change approaches. It's probably important to be pluggable as I"m not building k8s myself, rather, using the RHEL packages from Atomic upgrades.

Proposals:

• New files in docs/proposals

Major API changes:

• Changes to pkg/api/register.go or pkg/expapi/register.go (note that the latter will move soon kubernetes/kubernetes#13772)

Minor API changes:

• Changes to pkg/api/types.go or pkg/expapi/types.go

Will be useful if/when we start tracking PR latency

#112 refers

We should only merge PR's if Jenkins e2e tests are stable across multiple runs. If I'm reading the code correctly:

https://github.com/kubernetes/contrib/blob/023e2c05df5899bf9734b0d6d21fad0b0617fb45/submit-queue/e2e.go#L93

it deems a single successful test run per job to constitute a "stable" job. In theory, a single fluky success could cause all pending PR's to auto-merge, which is bad.

In practise, we have multiple test jobs being checked, so if all of them become flaky, their intermittent successes would all have to line up correctly for PR's to merge (which is statistically unlikely). But even so, I'd suggest adding a "number of consecutive successes required" config value per job. Faster, more reliable test jobs (e.g. e2e-gce-parallel, at 20 min and 90%+ reliability) might require, say, 4 consecutive successes, while slower jobs (e.g. scalability, at multiple hours per run), might require fewer.

For example
kubernetes/kubernetes#13393

My guess is the user in question is not in the e2e whitelist and thus e2e will only run when told to run. Open/close will do nothing.

Where is the e2e whitelist stored?

Because of its flakiness, the gce-autoscaling project was removed from the list of Jenkins projects that the merge-bot checks (PRs in waiting is growing). The failures needs to be investigated and the project added back to that list soon.

Right now we use the get_url to obtain most of the addon and configuration files from the Github content.

Example:

- name: Make sure the system services namespace exists
  get_url:
    url=https://raw.githubusercontent.com/GoogleCloudPlatform/kubernetes/master/cluster/saltbase/salt/kube-addons/namespace.yaml
    dest="{{ kube_config_dir }}/addons/"
    force=yes

We should have a mechanism to use files from the local system. Since the contrib folder is now outside the main repository this is a little more difficult since we can not assume file locations. It would be nice to have as little logic in the setup.sh script as well but might be the best solution?

I propose we keep the downloading versions of the files and have the option to use local files. We do this with packageManager and localBuild for the actual kubernetes binaries now. Using this same technique we have the config files.

Logging the bug to track ideas, suggestions and progress.

I had to make some changes to run this from my mac with a local repo build.

first, we may want to show how to push local binaries - its tricky to get the right path, b/c you need to use the dockerized build and specify amd64 iirc

You need to use run.sh to build dockerized linux binaries. Then, you have to do something like this...

  3 --- a/ansible/roles/master/defaults/main.yml
  4 +++ b/ansible/roles/master/defaults/main.yml
  5 @@ -1,3 +1,3 @@
  6  kube_master_insecure_port: 8080
  7 
  8 -localBuildOutput: ../../_output/local
  9 \ No newline at end of file
 10 +localBuildOutput: ~/Development/kubernetes/_output/dockerized/bin/linux

and

 13 --- a/ansible/roles/master/tasks/localBuildInstall.yml
 14 +++ b/ansible/roles/master/tasks/localBuildInstall.yml
 16  ---
 17  - name: Copy master binaries
 18    copy:
 19 -    src: "{{ localBuildOutput }}/go/bin/{{ item }}"
 20 +    src: "{{ localBuildOutput }}/amd64/{{ item }}"
 21      dest: /usr/bin/
 22      mode: 0755
 23    with_items:

also, i think we might want to update vagrant example file for openstack

and finally in my main.yml, this is what it looks like. Mainly, the os_auth_url was different... So we should update the example vagrant file

 24 diff --git a/ansible/roles/node/defaults/main.yml b/ansible/roles/node/defaults/main.yml
 29 -localBuildOutput: ../../_output/local
 30 +localBuildOutput: ~/Development/kubernetes/_output/dockerized/bin/linux
 31 os_username: jvyas
 32 os_password: 123123123123
 33 os_tenant: "ENG CTO Office"
 34 os_auth_url: "http://os1-public.osop.rhcloud.com:5000/v2.0/tokens"
 35 os_region_name: "OS1Public"
 36 os_ssh_key_name: "JPeerindex"
 37 os_flavor: "m1.small"
 38 os_image: "_OS1_Fedora-Cloud-Base-22-20150521.x86_64.qcow2"
 39 os_security_groups:
 40   - "default"
 41 os_floating_ip_pool: "os1_public"

Tracking bug to test and bump haproxy in service loadbalancer to 1.6, now that it's official (related #120, http://marc.info/?l=haproxy&m=144476286029747)

1. Make https://github.com/kubernetes/contrib/blob/master/Ingress/controllers/gce/rc.yaml look like https://github.com/kubernetes/kubernetes/tree/master/cluster/addons/cluster-loadbalancing/glbc.
2. Lower case the G on: https://github.com/kubernetes/contrib/blob/master/Ingress/controllers/gce/utils.go#L146 and https://github.com/kubernetes/contrib/blob/master/Ingress/controllers/gce/utils.go#L250
3. Change the default cluster name to "default-cluster-name" (https://github.com/kubernetes/contrib/blob/master/Ingress/controllers/gce/main.go#L59).

Hi,

I am trying to run the example, but I get stuck in "Expose services"; none of my nodes have any external IP. The following command gives me nothing:

kubectl get nodes -o json | grep -i externalip -A 1

How can I activate the external IP for the role=loadbalancer nodes?

Thank you

The submit queue automatically re-runs Jenkins before merging, but it's possible for the Shippable and Travis results to be very stale. Because Jenkins doesn't run all of the same verifications and checks currently, it's possible for things to break.

Motivating example: kubernetes/kubernetes#15608 was auto-merged, but it immediately broke the build, since some new docs had been added since its last CI run.

The GitHub API indicates when commit statuses were updated (see https://github.com/google/go-github/blob/master/github/repos_statuses.go#L34); it should hopefully be fairly straightforward to require these to be no more than N days old or retrigger CI if they are stale.

@brendandburns @eparis

Hello all,

I'm behind a proxy, so the first thing I had to solve was that contrib/ansible/roles/kubernetes/files/make-ca-cert.sh was not using the environment for a proxy nor a curlrc file. I ended up adding the --proxy my.proxy.here:3128 to the curl call directly. After that, the script had a problem with ./easyrsa --batch "--req-cn=${cert_ip}@$(date +%s)" build-ca nopass as ${cert_ip}@$(date +%s) would resolve to a string longer than 64 characters. After I worked around that, the ansible run went on up until

TASK: [master | Copy master binaries] *****************************************

which in turn failed due to:

fatal: [kube-master.WHOOPS] => input file not found at \ 
/home/WHOOPS/kubernetes/contrib/ansible/roles/_output/local/go/bin/kube-apiserver \
or /home/WHOOPS/kubernetes/_output/local/go/bin/kube-apiserver

And because I'm new to ansible I was wondering, because I can see this in roles/master/tasks/localBuildInstall.yml

- name: Copy master binaries
 copy:
   src: ../../_output/local/go/bin/{{ item }}
   dest: /usr/bin/
   mode: 755
 with_items:
   - kube-apiserver
   - kube-scheduler
   - kube-controller-manager
   - kubectl
 notify: restart daemons

making sense but prior to that it doesn't say that it'll build something. Nothing. So I'm curious, when should building the mentioned binaries in that _outcome directory happen?

So I was following

http://kubernetes.io/v1.0/docs/getting-started-guides/fedora/fedora_ansible_config.html

which already had the trouble of not being up2date (it doesn't state that the ansible stuff has moved) and that has lead so far nowhere.

Is this a cluster.yml switch, to enable that build?

All components that connect to etcd directly should support a secured etcd deployment. Podmaster only allows passing in the list of servers... it should also take a etcd client config file (like the API server)

https://github.com/kubernetes/contrib/blob/36816275fd53c7a2ef59650c80e2820fe3595584/pod-master/podmaster.go#L140

The documentation for the service loadbalancer indicates that the kube-ui can be reached via the default configuration however the loadbalancer does not query the kube-system namespace so nothing is actually exposed. Please either change the documentation to reflect this or allow a way to query multiple namespaces e.g kube-system, default, etc.

TASK: [flannel | Load the flannel config file into etcd] **********************
failed: [10.66.15.133 -> 10.66.15.133] => {"changed": true, "cmd": "/usr/bin/etcdctl --no-sync --peers=http://10.66.15.133:2379 set /cluster.local/network/config < /tmp/flannel-conf.json", "delta": "0:00:00.214135", "end": "2015-11-15 01:33:17.547932", "rc": 4, "start": "2015-11-15 01:33:17.333797", "warnings": []}
stderr: Error: 501: All the given peers are not reachable (failed to propose on members [http://10.66.15.133:2379] twice [last error: Put http://10.66.15.133:2379/v2/keys/cluster.local/network/config: dial tcp 10.66.15.133:2379: connection refused]) [0]

FATAL: all hosts have already failed -- aborting

As an example:

PR kubernetes/kubernetes#15014
was merged at 00:51 PST

at the same time "kubernetes-e2e-gce" was yellow

cc @lavalamp @eparis @brendandburns

maybe to prometheus? using something like https://github.com/prometheus/haproxy_exporter?

I'm trying to run through the ansible setup and keep getting to the following spot and it just hangs: Load the flannel config file into etcd

I'm running Fedora Server 22 from a Fedora client. Are there any additional logs I can see or things I can dig into to see why it's hanging?

/api says

json: unsupported type: <-chan struct {}

@lavalamp

The idea is to start from scratch guiding the user from 0 to k8s with the service loadbalancer configured
Something like https://gist.github.com/aledbf/2985931789f3c182f22b

@Beeps:

In newLoadBalacnerController lines https://github.com/kubernetes/contrib/blob/master/service-loadbalancer/service_loadbalancer.go#L391-L395

Seems a little off to me, i.e. line 391...

 lbc := loadBalancerController{ 
...
    tcpServices: map[string]int{},
}

In the above, shouldn't the loadBalancerController be binding tcpServices to the value of cfg.tcpServices ? The code just below seems to expect that tcpServices actually has some data in it...

strings.Split(*tcpServices, ",") {

I'm in the process of extracting more info from reading the code.

If I've missed something obvious, apologies in advance :)..

I'm trying out the openstack vagrant steps which seem to be working with only two small changes my side which I'll raise pr's for but then fail with the error below.

TASK: [opencontrail | Interface configuration file (physical)] ****************
fatal: [kube-node-2] => One or more undefined variables: 'opencontrail_interface' is undefined
fatal: [kube-node-1] => One or more undefined variables: 'opencontrail_interface' is undefined

FATAL: all hosts have already failed -- aborting

but as a total guess I uncommented https://github.com/kubernetes/contrib/blob/master/ansible/group_vars/all.yml#L50 and set it to eth0 which was an available interface - no idea if this was right at all but that error seemed to go away the next time I ran the vagrant up.

So now I'm left with the errors below which I guess are related and wondering if anyone has any suggestions

TASK: [opencontrail | build tag] **********************************************
fatal: [kube-master] => error while evaluating conditional: inventory_hostname in groups['nodes'] or inventory_hostname in groups['gateways']
ok: [kube-node-1]
ok: [kube-node-2]

TASK: [opencontrail | Unpack vrouter tarball] *********************************
changed: [kube-node-1]
changed: [kube-node-2]

TASK: [opencontrail | Depmod] *************************************************
changed: [kube-node-1]
changed: [kube-node-2]

TASK: [opencontrail | Reduce memory utilization of vrouter] *******************
skipping: [kube-node-1]
skipping: [kube-node-2]

TASK: [opencontrail | Interface up/down scripts] ******************************
changed: [kube-node-2] => (item=ifup-vhost)
changed: [kube-node-1] => (item=ifup-vhost)
changed: [kube-node-2] => (item=ifdown-vhost)
changed: [kube-node-1] => (item=ifdown-vhost)

TASK: [opencontrail | Interface configuration file (physical)] ****************
changed: [kube-node-2]
changed: [kube-node-1]

TASK: [opencontrail | Interface configuration file (vhost0)] ******************
fatal: [kube-node-2] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute 'ipaddr'", 'failed': True}
fatal: [kube-node-1] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute 'ipaddr'", 'failed': True}
fatal: [kube-node-2] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute 'ipaddr'", 'failed': True}
fatal: [kube-node-1] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute 'ipaddr'", 'failed': True}

FATAL: all hosts have already failed -- aborting

PLAY RECAP ********************************************************************
           to retry, use: --limit @/Users/jamesrawlings/cluster.retry

kube-master                : ok=92   changed=40   unreachable=1    failed=0
kube-node-1                : ok=63   changed=25   unreachable=1    failed=0
kube-node-2                : ok=63   changed=25   unreachable=1    failed=0

An unknown error happened in Vagrant OpenStack provider

To easily debug what happened, we recommend to set the environment
variable VAGRANT_OPENSTACK_LOG to debug

    $ export VAGRANT_OPENSTACK_LOG=debug

If doing this does not help fixing your issue, there may be a bug
in the provider. Please submit an issue on Github at
https://github.com/ggiamarchi/vagrant-openstack-provider
with the stracktrace and the logs.

We are looking for feedback, so feel free to ask questions or
describe features you would like to see in this provider.
Catched Error: Catched Error: Ansible failed to complete successfully. Any error output should be
visible above. Please fix these errors and try again.

I created kubernetes/kubernetes#12950 and didn't assign someone, and the mergebot later assigned me (probably because of the files I was touching).

It probably shouldn't self-assign PRs.

Perhaps make it wider? Or at least center it?

Building current servicelb and pushing it out to a cluster with the following services: https://gist.github.com/bprashanth/a1eaa27ef28f2c7fee0f using the rc.yaml in the servicelb directory, gives me the following errors in kubectl logs:

ALERT] 273/173127 (28) : parsing [/etc/haproxy/haproxy.cfg:87] : error detected while parsing switching rule : no such ACL : 'host_acl_kubernetes:443'.
[ALERT] 273/173127 (28) : parsing [/etc/haproxy/haproxy.cfg:92] : error detected while parsing switching rule : no such ACL : 'host_acl_echoheadersdefault'.
[ALERT] 273/173127 (28) : parsing [/etc/haproxy/haproxy.cfg:97] : error detected while parsing switching rule : no such ACL : 'host_acl_echoheadersx'.

And I can't reach http://lbip/echoheadersx, but I can reach the service through http://lbip:nodePort-of-echoheadersx. Guessing this is because of the latest host headers change.

@aledbf

reqrep and keep-alive do not play nicely. The first request to a http service works fine but subsequent requests 404 because the request is not parsed. Haproxy merely passes the data through. Using "option http-server-close" fixes this issue.

Issue noted here a few years back http://permalink.gmane.org/gmane.comp.web.haproxy/7670

Several useful fixes have gone in/are in flight/will go in soon:

• reaping zombie haproxy
• logging to syslog
• dry run mode talking to localhost kubectl proxy
• using alpine linux
• url based routing (maybe?)
• liveness probe chk directly in haproxy
• increasing relist period to 10m
• proper flag parsing

Need to test/push a new image to google_containers

We have the gce-e2e suite that acts as a gatekeeper for PR merges.

We should have the ability to add additional suite(s) to act as a gatekeeper for a PR merge.
This can be done by leveraging github labels.

Currently, we are in a mode where many tests are kicked out of gce-e2e suite because of slowness etc..

Having a custom label-driven suite gatekeeper, we can have certain PRs that need a specific test suite to be run, act as a gatekeep for the PR merges.

@quinton-hoole
@brendandburns
@ixdy
@thockin

https://github.com/kubernetes/contrib/blob/023e2c05df5899bf9734b0d6d21fad0b0617fb45/submit-queue/e2e.go#L93

As far as I can tell by looking at this code, this function returns true even if stable==false for some or all builds.

In theory this will cause PR's to be merged even if Jenkins e2e tests are failing.

@gmarek has reported this to be the case and @lavalamp has therefore disabled the merge bot.

Need to verify that this theoretical bug is in fact manifesting. The fix is pretty straightforward.

@brendandburns FYI.

Create a sidecar container that runs in the same pod as your backend and reports it's qps to a service annotation. The servicelb will lookup these service annotations and update the weights of the backends by writing to the haproxy socket (or #120). The goal here is to send less traffic to backends that have historically been slow to repond (so not exactly leastconns). I expect this to help with noisy neighbors.

@aledbf @jayunit100

From the logs right now.

I1006 20:33:38.846890       1 github.go:668] Fetching page 16510 of issues
I1006 20:33:40.737390       1 github.go:668] Fetching page 16511 of issues
I1006 20:33:42.847761       1 github.go:668] Fetching page 16512 of issues
I1006 20:33:44.840999       1 github.go:668] Fetching page 16513 of issues
I1006 20:33:46.855748       1 github.go:668] Fetching page 16514 of issues
I1006 20:33:48.763123       1 github.go:668] Fetching page 16515 of issues
I1006 20:33:50.816153       1 github.go:668] Fetching page 16516 of issues
I1006 20:33:52.949485       1 github.go:668] Fetching page 16517 of issues
I1006 20:33:54.800915       1 github.go:668] Fetching page 16518 of issues
I1006 20:33:56.796266       1 github.go:668] Fetching page 16519 of issues
I1006 20:33:58.795852       1 github.go:668] Fetching page 16520 of issues
I1006 20:34:00.960203       1 github.go:668] Fetching page 16521 of issues
I1006 20:34:02.973533       1 github.go:668] Fetching page 16522 of issues
I1006 20:34:05.166156       1 github.go:668] Fetching page 16523 of issues
I1006 20:34:06.867732       1 github.go:668] Fetching page 16524 of issues
I1006 20:34:08.911436       1 github.go:668] Fetching page 16525 of issues
I1006 20:34:10.828744       1 github.go:668] Fetching page 16526 of issues
I1006 20:34:12.900781       1 github.go:668] Fetching page 16527 of issues
I1006 20:34:14.876124       1 github.go:668] Fetching page 16528 of issues
I1006 20:34:16.764546       1 github.go:668] Fetching page 16529 of issues
I1006 20:34:18.790262       1 github.go:668] Fetching page 16530 of issues
I1006 20:34:20.993383       1 github.go:668] Fetching page 16531 of issues
I1006 20:34:22.817142       1 github.go:668] Fetching page 16532 of issues
I1006 20:34:24.834514       1 github.go:668] Fetching page 16533 of issues
I1006 20:34:26.800389       1 github.go:668] Fetching page 16534 of issues
I1006 20:34:28.815217       1 github.go:668] Fetching page 16535 of issues

We check LGTM before we start the merge testing process, but we don't double-check the LGTM. So if it is removed in the meantime, we won't notice.

I followed the step described in README.

I run into an issue of

core@kube-node-01 ~ $ docker logs c3e318d07f32
I0827 21:43:17.144454       1 service_loadbalancer.go:383] Creating new loadbalancer: {Name:haproxy ReloadCmd:./haproxy_reload Config:/etc/haproxy/haproxy.cfg Template:template.cfg Algorithm:roundrobin}
I0827 21:43:17.144510       1 service_loadbalancer.go:426] All tcp/https services will be ignored.
F0827 21:43:17.145153       1 service_loadbalancer.go:436] Failed to create client: open /var/run/secrets/kubernetes.io/serviceaccount/token: no such file or directory

I was using the same command kubectl create -f ./rc.yaml. The state of the pods show Running, but it goes into an infinite loop of Restart container.

These are two e2e not required prs:

kubernetes/kubernetes#14761
kubernetes/kubernetes#14494

Merge bot comments (multiple times):

Automatic merge from submit-queue

But the PRs are not merged.

cc @brendandburns @lavalamp @eparis

Currently is not possible to expose a service using HTTPS with a custom SSL certificate.

Under https://github.com/kubernetes/contrib/blob/master/ansible/roles/node/tasks/main.yml#L26 when you need to setup such that the client nodes can ssh back into the master node. Not exactly clear from the docs.

/cc @rrati

During the vagrant up I get an error as docker isn't running in the image during the opencontrail | Build docker container task, I've worked around it by adding service: name=docker state=started to https://github.com/kubernetes/contrib/blob/master/ansible/roles/docker/tasks/main.yml#L49 but I'm pretty sure my lack of understanding here means that's not the right answer.

Using fedora 22 cloud base image from https://download.fedoraproject.org/pub/fedora/linux/releases/22/Cloud/x86_64/Images/Fedora-Cloud-Base-22-20150521.x86_64.qcow2

TASK: [opencontrail | Build docker container] *********************************
skipping: [kube-node-1]
skipping: [kube-node-2]
failed: [kube-master] => {"changed": true, "cmd": ["docker", "build", "-t", "opencontrail/kmod_fedora22-4.0.4-301.fc22.x86_64", "fedora22-4.0.4-301.fc22.x86_64"], "delta": "0:00:00.023080", "end": "2015-10-20 18:01:26.362457", "rc": 1, "start": "2015-10-20 18:01:26.339377", "warnings": []}
stderr: Post http:///var/run/docker.sock/v1.20/build?cgroupparent=&cpuperiod=0&cpuquota=0&cpusetcpus=&cpusetmems=&cpushares=0&dockerfile=Dockerfile&memory=0&memswap=0&rm=1&t=opencontrail%2Fkmod_fedora22-4.0.4-301.fc22.x86_64&ulimits=null: dial unix /var/run/docker.sock: no such file or directory.
* Are you trying to connect to a TLS-enabled daemon without TLS?
* Is your docker daemon up and running?

Hi -

I've built and launched the servicelb container and launched it using the rc.yml example. I can connect to the stats port on 1936, and haproxy appears to be listening ok on 80. I've launched a test container to get it picked up by the loadbalancer but when I curl -v http://10.100.41.6/test I get 'Not Found'. If I query the logs for the loadbalancer I get the messages below. From here, I'm not sure where to go to try and further diagnose why it's not working but I'm guessing that the launched test container has not been detected so the haproxy.cfg has not been automatically updated (if that's correct?) - any help on how to try and get to bottom of this would be greatly appreciated.

Thanks,
Piers.

Icore@master ~ $ kubectl logs service-loadbalancer-1t2qm

I1110 05:58:36.959150 1 service_loadbalancer.go:550] Creating new loadbalancer: {Name:haproxy ReloadCmd:./haproxy_reload Config:/etc/haproxy/haproxy.cfg Template:template.cfg Algorithm: startSyslog:false lbDefAlgorithm:roundrobin}
E1110 05:58:36.959519 1 service_loadbalancer.go:241] Get : unsupported protocol scheme ""
I1110 05:58:36.988153 1 service_loadbalancer.go:294] haproxy -- cat: can't open '/var/run/haproxy.pid': No such file or directory