Erik Rose: WebLion Hosting—Leveraging laziness, impatience, and hubris
A talk at Plone Symposium East 2009. Track A, 11am, May 28
Join me on a whirlwind tour of WebLion's hosting service in a box!
Nobody at Penn State agrees on anything: what Plone version to run, what products to use, how to configure Apache, or what other fun things to run on the server. In the old days, when WebLion had only a few clients, I'd meet individually with the IT staff of each, help them set up their own servers, and, inevitably, spend hours remote-troubleshooting their heterogeneous setups (beset by firewalls and VPNs all the while).
With today's client list, I can't possibly pull this off and still get any coding done. Thus, WebLion Hosting was born: a way to keep a mob of heterogeneous Plone servers up to date with only O(1) effort.
I'll spill all the beans (and source code!):
- Packaging. How deployment and upgrades work, thanks to Debian's excellent Advanced Package Tool and MIT's config-package-dev configuration packaging framework.
- Philosophy. We handle unattended updates more reliably than buildout, which has a point of failure at every one of serveral servers and leaves installations destroyed when it fails halfway through. We allow for local configuration changes and don't pave over them with updates. We also leverage the packaging work and QA done by Debian, giving us a dependable stack without having to maintain our own build recipes for Apache, Squid, etc.
- Challenges. How egg-based Plone makes reliable deployment difficult and what we should do about it. How to use WebLion Hosting on your own machines.