Update status code for CRUD APIs

This commit updates the status code from 200 to 202 for create, update
or delete APIs.

Fixes: openfaas#220

Signed-off-by: Vivek Singh <[email protected]>

Expand verbs allowed for proxying

- Adds additional verbs as per g/w and watchdog

Signed-off-by: Alex Ellis (VMware) <[email protected]>

Chart version bumped to 1.2.3

- 1.2.3 version moves the gateway to 0.9.1

Signed-off-by: Alex Ellis (VMware) <[email protected]>

Add annotation support

Store stack.yml annotation meta data in deployment, pod, service and
replicaset

Relates to openfaas/faas#682

Signed-off-by: Edward Wilde <[email protected]>

Correct error in CI

The "latest-dev" tag was specified when it did not exist.

Signed-off-by: Alex Ellis <[email protected]>

Push image to both Quay.io and Docker Hub

With this change faas-netesd will be pushed to Quay.io
on release as a backup

Signed-off-by: Ivana Yovcheva (VMware) <[email protected]>

Bump gw/faas-netes + fix label update

Now that we have a label selector which is distinct from our own
labels we have to update the deployment labels in the update code.

Bumps GW to 0.8.2 (basic auth support) and faas-netesd to latest
for label enhancements. Tested on GKE 1.9.6.

Signed-off-by: Alex Ellis (VMware) <[email protected]>

Move label query to deployment.labels

When querying labels we were returning the selector labels instead
of the deployment labels. This change corrects the regression
which was introduced in 0.5.3.

Signed-off-by: Alex Ellis (VMware) <[email protected]>

Add Selector for deployment

A bug was found where updating through OpenFaaS-Cloud without a
faas_function label was causing an error with the buildshiprun
function. The label selector is required so we have an implicit
match and do not allow K8s to assume we want to match on all labels
because when it does - that means we can't change those labels in
an update.

Tested on kubeadm 1.9.6.

Signed-off-by: Alex Ellis (VMware) <[email protected]>

Fix 198 to optimize fetching single functions

An optimization is introduced to the single replica reader which
means we use "Get" rather than "List" and filtering in memory.
I originally wrote the List/filter approach with the intention of
applying this change later. @imikushin from VMware suggested this
change and raised PR196 - since there were some issues with
process, Kubernetes version used and (lack of) testing I've
implemented the change and done testing for this.

[x] List functions gives correct content-type and list
[x] Individual function queries return the correct data and
content-type
[x] Individual function queries for non-existant functions give
404 when not found

Also tested through the OpenFaaS UI including testing that
there was no regression for the readiness feature.

Signed-off-by: Alex Ellis (VMware) <[email protected]>