This changes introduce goreleaser, which does cross-compile and
package, as well as add version into go code. This change also
changes .travis.yml to allow to other users' forked repo.
This change introduces two fix about CNIVersion. One is to add CNIVersion into
delegated CNI conflist in case of DEL if CNIVersion is missing. ParseVersion() in libcni
checks cniVersion as mandatory field and sometimes user/daemonset miss it.
The latest flannel daemonset yaml (of flannel github) does not have CNIVersion, for example.
This change adds CNIVersion from multus config if cniVersion is missing (empty) in delegated
CNI config.
In addition, this fix also adds '--cni-version' in entrypoint.sh to add cniVersion in multus
config in case of '--multus-conf-file=auto'.
K8s with docker runtime, cmdDel() is invoked with empty network
namespace in case of restart node. Currently multus just returns
but CNI spec mention that we should invoke CNI plugin to cleanup.
This PR deletes "return nil" and proceeds to invoke CNI plugins
DEL. Fix#323.
This allows host-device plugin to recognize
Device PCI address passed from Multus.
It is related to the change in host-device which
enables use of device pci address as a config option:
https://github.com/containernetworking/plugins/pull/300
The flannel yaml in the example is using 10.244.0.0/16. If user bootstraps
a k8s cluster with different pod cidr, pod traffic will not go thru the flannel
vxlan tunnel.
This patch fixes the issue described in #289 where deviceID for
delegate plugin was not adding properly if the plugin conf inside
NetConfList.
Change-Id: I1d221f6b0e60a5b888b8e823611dfe12635e6897
Signed-off-by: Abdul Halim <abdul.halim@intel.com>
This patch fixes the issue described in #289 where deviceID for
delegate plugin was not adding properly if the plugin conf inside
NetConfList.
Change-Id: I1d221f6b0e60a5b888b8e823611dfe12635e6897
Signed-off-by: Abdul Halim <abdul.halim@intel.com>
With RollingUpdate update strategy, after you update a DaemonSet template, old DaemonSet pods will be killed, and new DaemonSet pods will be created automatically, in a controlled fashion.
This patch fixes the issue described in #289 where deviceID for
delegate plugin was not adding properly if the plugin conf inside
NetConfList.
Change-Id: I1d221f6b0e60a5b888b8e823611dfe12635e6897
Signed-off-by: Abdul Halim <abdul.halim@intel.com>
This change introduces kubelet client to get allocated device
information of a Pod from newly added Kubelet grpc service.
For more information please see:
[kubernetes/kubernetes#70508](https://github.com/kubernetes/kubernetes/pull/70508)
Change-Id: I11e58ccdd52662601f445fa24c7d55c225441efc
Signed-off-by: Abdul Halim <abdul.halim@intel.com>
(cherry picked from commit d3c92b4aa2)
adding new imported package dependencies in vendor which is required
for Kubelet Pod Resource api client.
Change-Id: If6c74598e12af5f8659df69371e72dd064823f49
(cherry picked from commit 8ee7eb335e)
This change introduces kubelet client to get allocated device
information of a Pod from newly added Kubelet grpc service.
For more information please see:
[kubernetes/kubernetes#70508](https://github.com/kubernetes/kubernetes/pull/70508)
Change-Id: I11e58ccdd52662601f445fa24c7d55c225441efc
Signed-off-by: Abdul Halim <abdul.halim@intel.com>
(cherry picked from commit d3c92b4aa2)
adding new imported package dependencies in vendor which is required
for Kubelet Pod Resource api client.
Change-Id: If6c74598e12af5f8659df69371e72dd064823f49
(cherry picked from commit 8ee7eb335e)
This change introduces kubelet client to get allocated device
information of a Pod from newly added Kubelet grpc service.
For more information please see:
[kubernetes/kubernetes#70508](https://github.com/kubernetes/kubernetes/pull/70508)
Change-Id: I11e58ccdd52662601f445fa24c7d55c225441efc
Signed-off-by: Abdul Halim <abdul.halim@intel.com>
adding new imported package dependencies in vendor which is required
for Kubelet Pod Resource api client.
Change-Id: If6c74598e12af5f8659df69371e72dd064823f49
This fixes#243 with following changes:
+ Optimize to fetch Pod from k8s client
+ Change to use cache always in DEL.
+ If failed to fetch the pod info from k8s clinet in deletion,
use cached delegates as emergency bailout
+ Add test cases for cache
fixed conflicts in multus/multus.go & multus/multus_test.go
