The redis example in the dev quickstart needs update/clarification

[Martin Falatic]

Considering dev-quickstart.rst#exercising-the-services-using-devstack:

The last step appears to be broken:

ssh minion@$(nova list | grep 10.0.0.4 | awk '{print $13}')
REDIS_ID=$(docker ps | grep redis:v1 | grep k8s_redis | tail -n +2 | awk '{print $1}')
docker exec -i -t $REDIS_ID redis-cli

...

There are four redis instances, one master and three slaves, running across the bay, replicating data between one another.”

Behavior and workaround as reported by user:

I see one master redis instance on one minion and one slave redis instance on a second minion (each has its own associated sentinel container as expected).
The redis-cli command times out with “Could not connect to Redis at 127.0.0.1:6379: Connection refused”. HOWEVER, if I add a host IP and port for the redis master minion (“-h 10.100.84.2 -p 6379”), the example works.

A workaround, using “-h 10.100.84.2 -p 6379“:

[minion@k8-4gmqfvntvm-0-6fymzzw3wrjx-kube-minion-zjdejo5sffxv ~]$ REDIS_ID=$(sudo docker ps | grep redis:v1 | grep k8s_redis | awk '{print $1}')
[minion@k8-4gmqfvntvm-0-6fymzzw3wrjx-kube-minion-zjdejo5sffxv ~]$ sudo docker exec -i -t $REDIS_ID redis-cli -h 10.100.84.2 -p 6379
10.100.84.2:6379> get replication:test
"true"
10.100.84.2:6379>

Note: address for the redis master was found by running the following on the master minion:

sudo docker exec -i -t $REDIS_ID ip addr show dev eth0

5: eth0: <BROADCAST,UP,LOWER_UP> mtu 1472 qdisc noqueue state UP
    link/ether 02:42:0a:64:54:02 brd ff:ff:ff:ff:ff:ff
    inet 10.100.84.2/24 scope global eth0
       valid_lft forever preferred_lft forever
    inet6 fe80::42:aff:fe64:5402/64 scope link
       valid_lft forever preferred_lft forever

The documentation should be updated to reflect this workaround if it is indeed the right way to do this now.

[Martin Falatic (martinfalatic)]

Currently getting an error during pod-create (which used to work):
ERROR: 'NoneType' object has no attribute 'status' (HTTP 500)

May need to update docs to reflect newer k8s version as well? Not sure yet.

e.g., https://github.com/GoogleCloudPlatform/kubernetes/releases/download/v1.0.1/kubernetes.tar.gz

Still throws a (different) error:

ERROR: Timed out waiting for a reply to message ID 5c9fa2ec09224f6bae1c1b2da2a6252d (HTTP 500)

[Martin Falatic (martinfalatic)]

Update: at this point, the goal is to add clarity concerning how many bay nodes are really necessary for this test (one master + one minion will suffice), what things look like on the minion before and after instantiation of redis (via docker ps), and clarifying the "log into another container host" bit (which is only relevant if you have more than one bay node and want to test that configuration).

It's not that it's broken at this point, it's just rather confusing to the user.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/232434

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/232434
Committed: https://git.openstack.org/cgit/openstack/magnum/commit/?id=5632553073dcf038283b6891533d74f28f574b08
Submitter: Jenkins
Branch: master

commit 5632553073dcf038283b6891533d74f28f574b08
Author: Martin Falatic <email address hidden>
Date: Thu Oct 8 02:18:19 2015 -0700

    Update and clarify redis examples in quickstart

    Update and clarify redis examples in quickstart guide.

    DocImpact
    Closes-Bug: #1479115

    Change-Id: Icbf0ddd501a60c4369ce3431f5260a593f23f078