OpenStack Affinity and Anti-Affinity

Server groups with affinity and anti-affinity policies control how the Nova scheduler places instances across compute hosts. Affinity keeps instances together on the same host; anti-affinity spreads them across different hosts. This is essential for high availability and performance optimization.

When to Use Each Policy

Prerequisites

Step 1: Enable the Scheduler Filters

Verify the required filters are enabled in /etc/nova/nova.conf on the controller:

[filter_scheduler]
enabled_filters = AvailabilityZoneFilter,ComputeFilter,ComputeCapabilitiesFilter,ImagePropertiesFilter,ServerGroupAntiAffinityFilter,ServerGroupAffinityFilter

Restart Nova scheduler:

sudo systemctl restart nova-scheduler

Step 2: Create an Anti-Affinity Server Group

Anti-affinity ensures each instance lands on a different compute host:

openstack server group create \
  --policy anti-affinity \
  ha-web-group

Note the group ID from the output.

Step 3: Launch Instances in the Group

openstack server create \
  --flavor m1.medium \
  --image "Ubuntu 22.04" \
  --network tenant-net \
  --hint group=<ha-web-group-id> \
  web-01

openstack server create \
  --flavor m1.medium \
  --image "Ubuntu 22.04" \
  --network tenant-net \
  --hint group=<ha-web-group-id> \
  web-02

openstack server create \
  --flavor m1.medium \
  --image "Ubuntu 22.04" \
  --network tenant-net \
  --hint group=<ha-web-group-id> \
  web-03

Each instance will be placed on a different compute host. If there are not enough hosts, the launch fails (use soft-anti-affinity to relax this).

Step 4: Verify Placement

openstack server list --long -c Name -c Host

Expected output:

+--------+------------+
| Name   | Host       |
+--------+------------+
| web-01 | compute-01 |
| web-02 | compute-02 |
| web-03 | compute-03 |
+--------+------------+

Step 5: Create an Affinity Server Group

Affinity forces all instances onto the same host:

openstack server group create \
  --policy affinity \
  colocated-group

openstack server create \
  --flavor m1.small \
  --image "Ubuntu 22.04" \
  --network tenant-net \
  --hint group=<colocated-group-id> \
  app-01

openstack server create \
  --flavor m1.small \
  --image "Ubuntu 22.04" \
  --network tenant-net \
  --hint group=<colocated-group-id> \
  app-02

Both instances land on the same compute host.

Step 6: Use Soft Policies

Soft policies are best-effort—they prefer the policy but do not fail if it cannot be satisfied:

# Prefer different hosts, but allow colocation if needed
openstack server group create \
  --policy soft-anti-affinity \
  flexible-ha-group

# Prefer same host, but allow spread if needed
openstack server group create \
  --policy soft-affinity \
  flexible-colocated-group

Real-World Patterns

Database HA Cluster

openstack server group create --policy anti-affinity db-cluster
for i in 1 2 3; do
  openstack server create --flavor m1.xlarge \
    --image "Ubuntu 22.04" --network db-net \
    --hint group=<db-cluster-id> db-node-$i
done

Application + Cache Colocation

openstack server group create --policy affinity app-cache
openstack server create --flavor m1.medium \
  --image "Ubuntu 22.04" --network app-net \
  --hint group=<app-cache-id> app-server
openstack server create --flavor m1.small \
  --image "Ubuntu 22.04" --network app-net \
  --hint group=<app-cache-id> redis-cache

Managing Server Groups

# List all server groups
openstack server group list

# Show group details and members
openstack server group show <group-id>

# Delete a group (instances are not affected)
openstack server group delete <group-id>

Troubleshooting

Summary

Server groups with affinity and anti-affinity policies give you control over instance placement. Use anti-affinity for high availability across hosts, affinity for low-latency colocation, and soft variants when flexibility is needed.