Hacker News new | past | comments | ask | show | jobs | submit | _9lfj's comments login

Is it still the case that CF requires a ton of overhead to get running? Last I checked there was no easy way to set up CF without:

- Running on a cloud

- Spinning up ~8 different servers, each with ~2GB of memory iirc

Kinda a lot of overhead if you're just running a few dev environments like I was.


I get this question a lot.

PCFDev[0] if you just want a dev environment to tinker with. You can also tinker with cflocal[1] (unsupported) or BOSH-lite[2] (somewhat-supported) if you're looking to practice ops as well.

A lot of Cloud Foundry teams do integration testing with BOSH-lite on large single VMs, because it essentially presents BOSH with a "VM" interface that spins up nested containers instead.

You can, in theory, deploy however few or however many real VMs, or bare metal servers, as you wish, by editing a BOSH manifest.

By default the cf-deployment[3] manifest uses something like 23 VMs of varying sizes. That's because the default manifest is intended for production use with multi-AZ HA as the minimum level of reliability. If you don't need that, one of the alternatives I noted will get you what you want.

[0] https://pivotal.io/pcf-dev

[1] https://github.com/sclevine/cflocal

[2] https://github.com/cloudfoundry/bosh-lite

[3] https://github.com/cloudfoundry/cf-deployment


All the options in the article can be self-hosted or run anywhere you like, ECS is proprietary and Amazon only, which is likely why it was excluded. Heroku wasn't in the comparison either.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: