ocurrent / ocurrent-deployer

A pipeline that deploys unikernels and other services
23 stars 18 forks source link

Deploy staging version of the solver-service #200

Closed talex5 closed 1 year ago

talex5 commented 1 year ago

@tmcgilchrist what's the next step for getting this deployed?

I suggest:

  1. Merge this PR, allowing the new version to build.
  2. Configure the new solver (already on arm64-jade-1) to use the production pool rather than the staging one.
  3. Run the new service alongside the old one (on jade-2) for a bit. If the new one crashes, the old solver will take over. If the new version produces wrong results, stop it and let the old one handle all requests.
  4. Assuming it works, merge https://github.com/ocurrent/solver-service/pull/71 and shut down the old solver.

The /etc/systemd/system/solver-worker.service file I'm using with it is:

[Unit]
Description=Solver Worker
After=network.target

[Service]
ExecStart=/usr/local/bin/solver-service run-cluster -c /etc/ocluster/solver.cap --name='arm64-jade-1.equinix.ci.dev' --cache-dir=/var/cache/solver --capacity=8 --internal-workers=4 --verbosity=info
Restart=always

[Install]
WantedBy=multi-user.target

Note that the old service accidentally cloned Git repositories to the root. You can delete /github_com after switching to the new service.

talex5 commented 1 year ago

@mtelvers sounds good! I've updated the deployer and installed the binary that it produced on arm64-jade-1.

rikusilvola commented 1 year ago

Please remember to make a post about this on infra blog! If all goes well, it'll be good publicity for Eio.