Startup Semaphore

[jonwinton]

Quick semaphore link

Problem

In a production environment there are generally multiple Clay instances running at one time, whether it be multiple servers/containers or even a cluster on a single instance. At startup time Clay bootstraps data for components/sites, something we only need/want to run on one individual instance in a cluster/deployment group.

Proposal

Add semaphore support to the storage API, allowing whichever storage module is chosen to dictate how the data is stored. The data we might send across:

• Which node has begun bootstrapping. Could be determined with the os module using a hostname/pid combination to generate the key.
• Which deploy is being handled. Can we rely on the package.json version of the clay instance? Multiple environments are not always going to see this increment
• The state of the startup process: Not started, running, finished, failed? 0, 1, 2, 3?

Questions

• What happens if the startup process fails and needs to be restarted?
• Should plugins be able to use the same semaphore data to determine how they startup? For example, amphora-search does not ALWAYS need to check for the presence of indices on every Clay instance.

[amelvisfranco]

Semaphora

[amelvisfranco]

Semaphora