Cannot create a cluster with instances using a specific volume type because there is a mismatch between client and server code in referencing the type data in the request

[Janice-Chen]

Trove cluster-create service can not get the 'volume_type' value even specifies the volume_type when create a cluster using 'trove cluster-create' command.

Test Cli View:

trove cluster-create ClusterName mongodb mongodb-3.2 \
--instance "flavor=flavorId,volume=10,volume_type=ssd,nic='net-id=NetID'" \
--instance "flavor=flavorId,volume=10,volume_type=ssd,nic='net-id=NetID'" \
--instance "flavor=flavorId,volume=10,volume_type=ssd,nic='net-id=NetID'"

The bug is a clear error in the code.I found that python-troveclient describe volume_type in its request body with the key 'type' while trove cluster_create cli parsing with the key "volume_type".cluster_create cli will never get the specified volume_type in current codes.

This bug also affects bug:https://bugs.launchpad.net/trove/+bug/1622957，but can not resolve it.

[Amrith Kumar (amrith)]

How is this different from https://bugs.launchpad.net/trove/+bug/1622957

[Amrith Kumar (amrith)]

Please don't mark the bug in progress if it is a duplicate of the other almost identical bug.

[Janice-Chen (784059080-c)]

The two bugs are not the same, I submit another one, but the two are related indeed.For bug1 -https://bugs.launchpad.net/trove/+bug/1622957，it's mongodb who neglect volume_type parameter, but this bug will causes all datastore failed to get volume_type when creating a cluster with a specified volume type

[Amrith Kumar (amrith)]

your bugs really need to have more information about the problem you are facing. Something which we can use to assess how serious of a problem this is and triage the bug accordingly.

So please DO NOT mark the bug as in progress till it is fully documented.

[Amrith Kumar (amrith)]

Janice Chen has submitted a patch for this issue https://review.openstack.org/#/c/369873/1. The change seems to point to a real problem.

The client passes in volume_type as the parameter 'type' in the request however the server code for clusters appears to be looking for 'volume_type'.

I'm updating the bug to make it more informative. Here is the proposed fix.

diff --git a/trove/cluster/service.py b/trove/cluster/service.py
index 0cb7afc..e315fc8 100644
--- a/trove/cluster/service.py
+++ b/trove/cluster/service.py
@@ -160,7 +160,7 @@ class ClusterController(wsgi.Controller):
             volume_size = volume_type = nics = availability_zone = None
             if 'volume' in node:
                 volume_size = int(node['volume']['size'])
- volume_type = node['volume'].get('volume_type')
+ volume_type = node['volume'].get('type')
             if 'nics' in node:
                 nics = node['nics']
             if 'availability_zone' in node:

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/369873
Committed: https://git.openstack.org/cgit/openstack/trove/commit/?id=72330ac568e140a6d5fa3820a30ce61ec65a0b82
Submitter: Jenkins
Branch: master

commit 72330ac568e140a6d5fa3820a30ce61ec65a0b82
Author: chenshujuan <email address hidden>
Date: Wed Sep 14 15:38:29 2016 +0800

    Cluster code looks for wrong field in request for volume type

    The cluster code on the server side is looking for a field called
    'volume_type' in the request. The correct field should be 'type'.

    Closes-Bug: #1623005
    Change-Id: I5f46b887e028819dcd2d18da6390d38401528194
    Author: chenshujuan <email address hidden>
    Owner: chenshujuan <email address hidden>