Add persistent storage for CM state information

Issue-ID: DCAEGEN2-990
Change-Id: I122e541d6ea0fa6bca06157d6ae7a330048d2ed7
Signed-off-by: Jack Lucas <jflucas@research.att.com>

1 files changed, 31 insertions, 0 deletions

diff --git a/cm-container/README.md b/cm-container/README.md
index a29423d..6e1e26e 100644
--- a/cm-container/README.md
+++ b/cm-container/README.md
@@ -31,3 +31,34 @@ In a Kubernetes environment, we expect that the <path_to_kubeconfile_file> and t
 
 We also expect that in a Kubernetes environment the external port mapping would not be
 needed.
+
+## Persistent Storage
+In an ONAP deployment driven by OOM, Cloudify Manager will store data related to its state
+in a Kubernetes PersistentVolume.  If the Cloudify Manager pod is destroyed and recreated, 
+the new instance will have all of the state information from the previous run.
+
+To set up persistent, we replace the command run by the container (`CMD` in the Dockerfile) with
+our own script `start-persistent.sh`.  This script checks to see if a persistent volume has been
+mounted in a well-known place (`/cfy-persist` in the container's file system).  If so, the script 
+then checks to see if the persistent volume has been populated with data.  There are two possibilities:
+1. The persistent volume hasn't been populated, indicating that this is the first time Cloudify Manager is
+being run in the current environment.  In this case, the script copies state data from several directories in
+the container file system into directories in the persistent volume.  This is data (such as database schemas for
+Cloudify Manager's internal postgres instance) that was generated when the original Cloudify Manager image was
+created by Cloudify.
+2. The persistent volume has been populated, indicating that this is not the first time Cloudify Manager is being
+run in the current environment.   The data in the persistent volume reflects the state that Cloudify Manager was in
+when it exited at some point in the past.   There's no need to copy data in this case.
+In either case, the script will create symbolic links from the original data directories to the corresponding directories
+in the persistent store.
+
+If there is no persistent volume mounted, the script does nothing to set up persistent data, and the container will have
+no persistent storage.
+
+The last command in the script is the command from the original Cloudify version of the Cloudify Manager image. It runs `/sbin/init`,
+which then brings up the many other processes needed for a working instance of Cloudify Manager.
+
+## The `setup-secret.sh` script
+When Kubernetes starts a container, it mounts a directory containing the credentials that the container needs to access the Kubernetes API on the local Kubernetes cluster.  The mountpoint is `/var/run/secrets/kubernetes.io/serviceaccount`.   Something about the way that Cloudify Manager is started (possibly because `/sbin/init` is run) causes this mountpoint to be hidden.   `setup-secret.sh` will recreated the directory if it's not present and symbolically link it to a copy of the credentials mounted at `/secret` in the container file system.  This gives Cloudify Manager the credentials that the Kubernetes plugin needs to deploy Kubernetes-based DCAE components.
+
+`setup-secret.sh` needs to run after '/sbin/init'.  The Dockerfile installs it in the `rc.local` script that runs at startup.
\ No newline at end of file