Bug 1425706

Ge Liu what's the contents of your /tmp/config file? It looks like it is not complete cert.

Maciej, the /tmp/config created by step 1. above: #oc login ec2-52-xx-xx-xx.compute-1.amazonaws.com:443 --config=/tmp/config 


# cat /tmp/config
apiVersion: v1
clusters:
- cluster:
    insecure-skip-tls-verify: true
    server: https://ec2-xx-xx-1xx-xx.compute-1.amazonaws.com:8443
  name: ec2-52-90-xx-xx-compute-1-amazonaws-com:8443
contexts:
- context:
    cluster: ec2-52-90-xx-xx-compute-1-amazonaws-com:8443
    user: geliu/ec2-52-xx-xx-xx-compute-1-amazonaws-com:8443
  name: /ec2-52-90-xx-xx-compute-1-amazonaws-com:8443/geliu
- context:
    cluster: ec2-52-90-xx-xx-compute-1-amazonaws-com:8443
    namespace: lgproj
    user: geliu/ec2-52-xx-xx-213-compute-1-amazonaws-com:8443
  name: lgproj/ec2-52-90-1xx-xx-compute-1-amazonaws-com:8443/geliu
current-context: lgproj/ec2-52-xx-xx-xx-compute-1-amazonaws-com:8443/geliu
kind: Config
preferences: {}
users:
- name: geliu/ec2-52-90-xx-xx-compute-1-amazonaws-com:8443
  user:
    token: PDIs3IoVHgHWTXVBmkhj7bHStNwUiKfRi4lF8287UeY

Master PR: https://github.com/openshift/origin/pull/13072
1.5 PR: https://github.com/openshift/origin/pull/13073

Commits pushed to master at https://github.com/openshift/origin

https://github.com/openshift/origin/commit/174e3dd2174a1289700b3efabc59b5463a7a8141
Bug 1425706 - protect from nil tlsConfig.

In very special case (client config does not contain ca cert nor auth cert nor
set tls-insecure), tlsConfig can be nil and no error is returned from TLSConfigFor
either, thus we need to check for nil on tlsConfig before proceeding in
image pruning.

https://github.com/openshift/origin/commit/a6633738e69d433736c6677ee5931bad47ef8ea5
Merge pull request #13072 from soltysh/bug1425706

Merged by openshift-bot

Verified in ocp 3.5:

# openshift version
openshift v3.5.0.34
kubernetes v1.5.2+43a9be4
etcd 3.1.0

# oadm prune  images --all=true --registry-url=https://172.xx.xx.xx:5000 --certificate-authority=/tmp/config
Dry run enabled - no modifications will be made. Add --confirm to remove images
[root@host-8-172-x master]# oadm prune  images --all=true --registry-url=https://172.xx.xx.xx:5000 --certificate-authority=/tmp/config --confirm
[root@host-8-172-x master]# 

The panic error fixed, but it looks like the the image data still exist after run the prune command, I'm not sure whether the command line above it correct？


-Before run prune:

# oc rsh docker-registry-5-6c69f ls /registry/docker/registry/v2/repositories/lgproj/origin-ruby-sample
_layers  _manifests  _uploads

# oc rsh docker-registry-5-6c69f ls /registry/docker/registry/v2/repositories/lgproj/origin-ruby-sample/_manifests/revisions/sha256
091a11048fd3e796d4501b0c473bce09074ee39abe51c275ea4169f537c88b92

-After run prune:

# oc rsh docker-registry-5-6c69f ls /registry/docker/registry/v2/repositories/lgproj/origin-ruby-sample
_layers  _manifests  _uploads

# oc rsh docker-registry-5-6c69f ls /registry/docker/registry/v2/repositories/lgproj/origin-ruby-sample/_manifests/revisions/sha256
091a11048fd3e796d4501b0c473bce09074ee39abe51c275ea4169f537c88b92

@maszulik, could u pls take a look the issue in my comment 5? anyway, the panic issue have been fixed, so close this bug.

@geliu, there are multiple options why this layer was not removed, ie. this could be a shared layer with other image (eg. the builder), or it was not even considered eligible for pruning. It's hard to say, please perform thorough tests (increasing loglevel to see what's being removed helps here) and open a bug if you feel like there's a problem with pruning.

Since this bug never reached customers, I am closing it.