| 2016-08-11 15:54:01 |
Ilya Kharin |
description |
The cluster_upgrade extension does not support multi-rack installations and does not properly handle resources from different node groups.
Steps to reproduce:
1. Install 8.0 environment with nodes assigned to multiple node groups:
default: 3 x compute nodes
2 x ceph-osd nodes
r10: 3 x controller nodes
2. Upgrade master node to 9.0
3. Create upgrade environment:
upgrade-env
4. Upgrade primary controller:
upgrade-node
Expected results:
3. The new environment has the same set of node groups (default and r10) and the same set of networks per each node group.
4. The primary controller after upgrade is assigned in the respective node group (r10) and all interfaces are assigned in respective networks for the r10 node group, a set of VIPs are the same in respective networks for the r10 node group.
Actual result:
3. The new environment contains only the default node group with appropriate networks.
4. The primary controller is assigned in the default node group not r10 and its interfaces are assigned in networks that are belong to the default node group. VIps are assigned to networks of the default node group not the r10 node group.
Reproducibility:
Constantly.
Workaround:
3. After the upgrade-env command create appropriate set of node groups and their networks via UI/CLI.
4. Run upgrade-node with --no-provision flag, change a node group and an assignment of interfaces via UI/CLI, create VIPs vis CLI. |
The cluster_upgrade extension does not support multi-rack installations and does not properly handle resources from different node groups.
Steps to reproduce:
1. Install 8.0 environment with nodes assigned to multiple node groups:
default: 3 x compute nodes
3 x ceph-osd nodes
r10: 3 x controller nodes
2. Upgrade master node to 9.0
3. Create upgrade environment:
upgrade-env
4. Upgrade primary controller:
upgrade-node
Expected results:
3. The new environment has the same set of node groups (default and r10) and the same set of networks per each node group.
4. The primary controller after upgrade is assigned in the respective node group (r10) and all interfaces are assigned in respective networks for the r10 node group, a set of VIPs are the same in respective networks for the r10 node group.
Actual result:
3. The new environment contains only the default node group with appropriate networks.
4. The primary controller is assigned in the default node group not r10 and its interfaces are assigned in networks that are belong to the default node group. VIps are assigned to networks of the default node group not the r10 node group.
Reproducibility:
Constantly.
Workaround:
3. After the upgrade-env command create appropriate set of node groups and their networks via UI/CLI.
4. Run upgrade-node with --no-provision flag, change a node group and an assignment of interfaces via UI/CLI, create VIPs vis CLI. |
|
