How to add additional agents to mesosphere dc/os cluster?

Question

We have Mesosphere DC/OS version 1.7 running on 6 vmware vm's with rhel7 and we would now like to add more agents. We were able to increase from our initial 2 agents(vms) to 6 agents(vms) by running dcos_generate_confg.sh --uninstall from the boot node. We then added the new ip addresses to the agent list in genconf/config.yaml and ran --genconf, --preflight, --deploy, --postflight. This worked however the --uninstall blew away our existing public agent, which we had to go back in and re-install.

Is there another way to add agents without running the --uninstall? Very nervous about what that's doing under the hood. Thank you!

score 7 · Accepted Answer · edited May 04 '18 at 00:02

For my part, I also had trouble understanding how to add a node to a cluster, then I found this link that allowed me to do this:

The Mesosphere guide to getting started with DC/OS: Part 2

Go to the "Adding an internet-facing HAproxy load balancer" chapter.

To summarize what I did and which worked:

BootStrap Node:

sudo vi ~/genconf/config.yaml ## add node slave IP

sudo ./dcos_generate_config.sh --install-prereqs

sudo ./dcos_generate_config.sh --preflight

You can ignore errors from the existing agents and masters—DC/OS is just telling you it’s already installed, and you probably don’t want to clobber that installation in-place.

sudo scp -r -i genconf/ssh_key genconf/serve $NEW_NODE_IP:~ ## Copy to the node

New node:

sudo mkdir /opt/dcos_install_tmp

sudo cp -r serve/* /opt/dcos_install_tmp/

cd /opt/dcos_install_tmp/
chmod +x dcos_install.sh

Agent Private

     sudo ./dcos_install.sh slave

Agent Public

     sudo ./dcos_install.sh slave_public

BootStrap node:

sudo ./dcos_generate_config.sh --postflight

I hope that this can help you!

Thanks Benjamin, that worked great, just what we needed! A couple of notes .. on the `cp server/* /opt/dcos_install_tmp/` step, I had to use `cp -r `, there are a couple of required sub directories that were not picked up. Also, had to `chmod +x dcos_install.sh` and fetch_packages.sh. Just nits. The command above makes a public node. To make a private one use: `sudo ./dcos_install.sh slave`. To make a master node use: `sudo ./dcos_install.sh master`. We'll be trying that next. Thanks again. — grayzzz, Aug 05 '16 at 13:44

How to add additional agents to mesosphere dc/os cluster?

1 Answers1

You can ignore errors from the existing agents and masters—DC/OS is just telling you it’s already installed, and you probably don’t want to clobber that installation in-place.