production-ready-cluster.md

<- Back to index

WARNING: Currently deprecated since we have to use bootstrap to initialize a cluster.

Creating a production-ready cluster

After reviewing some CoreOS architectures we finally edited this to respond to architecture changes needs.

As so, we added a role option to specify if node is part of cluster boss providers or just a worker.

The example we will provide to you on this guide consists into:

• 4 service nodes (2-per-location)
• 10 worker nodes (5-per-location)

Locations will be:

• North Europe
• West US

If you have many accounts, feel free to spread nodes on them by supplying account number, id or random.

# WARNING:
# Please be sure that you don't have any nodes configured.
# You can do so by checking `coreos node list`


# building up service nodes
coreos node create \
    --provider="Azure" \
    --numberOfNodes=2 \
    --role="boss" \
    --location="North Europe" \
    --size="Medium" \
    --debug # Just because I like to see things happening :)

coreos node create \
    --provider="Azure" \
    --numberOfNodes=2 \
    --role="boss" \
    --location="West US" \
    --size="Medium" \
    --debug # Just because I like to see things happening :)

# building workers
coreos node create \
    --provider="Azure" \
    --numberOfNodes=5 \
    --role="worker" \
    --location="North Europe" \
    --size="Small" \
    --debug # Just because I like to see things happening :)

coreos node create \
    --provider="Azure" \
    --numberOfNodes=5 \
    --role="worker" \
    --location="West US" \
    --size="Small" \
    --debug # Just because I like to see things happening :)

# At this moment cluster should be up and running
# Since metadata is automatically built, you could configure units to run
# only on a specified location, on certain kind of nodes, and so on...
# Note: run `coreos fleetctl -- list-machines` for further meta details

# Now you can check up nodes by sshing into them
# coreos node list
# coreos node ssh -n NODEID

Scalling

You can scale cluster by:

• location, by adding or removing boss nodes on different locations;
• processing power, by adding or removing worker nodes on each location;