[HEAT] Support anti-affinity for node processes in many node groups during scaling
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Sahara |
Fix Released
|
Medium
|
Andrew Lazarev |
Bug Description
Heat engine doesn't allow users to define clusters with anti-affinity enabled on node process shared across many node groups.
For example the following cluster structure will not be scaled properly or scaled with error:
{
"name": "cluster",
"plugin_name": "vanilla",
"hadoop_
"anti_
"node_groups": [
{
"name": "master",
},
{
"name": "worker1",
},
{
"name": "worker1",
}
]
}
So in the above cluster structure "datanode" processes will not be scaled (up or down) properly after cluster provisioning.
The reason of that are nova scheduler hints definitions in heat templates which should be updated after each scaling operations. At the same time heat template already contains scheduler hints for provisioned instances in cluster with anti-affinity enabled and newly generated template during scaling leads to remove/create instances which are not dedicated to be removed/added.
Changed in savanna: | |
importance: | High → Medium |
Changed in savanna: | |
milestone: | icehouse-3 → next |
summary: |
- Support anti-affinity for node processes in many node groups during - scaling + [HEAT]Support anti-affinity for node processes in many node groups + during scaling |
summary: |
- [HEAT]Support anti-affinity for node processes in many node groups + [HEAT] Support anti-affinity for node processes in many node groups during scaling |
Changed in sahara: | |
milestone: | next → juno-1 |
Changed in sahara: | |
milestone: | juno-1 → juno-2 |
Changed in sahara: | |
milestone: | juno-2 → juno-3 |
Changed in sahara: | |
assignee: | nobody → Andrew Lazarev (alazarev) |
Changed in sahara: | |
status: | Fix Committed → Fix Released |
Changed in sahara: | |
milestone: | juno-3 → 2014.2 |
Fix proposed to branch: master /review. openstack. org/112159
Review: https:/