Support matching collections in the filesystem in addition to .config.php files during the config phase of loading a class. A .config.php file will be loaded first if found, followed by all files in the matching collection in lexical order.

Caching 404 from parent sites currently uses a global cache key namespace to gain some efficiency from multiple sites on the same server sharing the status. This however puts the cache key outside the scope of tools for clearing the VFS or application cache for a site, requiring a full PHP restart to clear it.

Prefix the cache key the same way that file/collection caches are keyed, so that the "repair filesystem" task's clear cache action will clear out 404 caches too

Observed in Firefox 42.

Investigated a bit and found that the data is passing through alright but the Emergence instances isn't recognizing the data as json.

• Set CURLOPT_CONNECTTIMEOUT' in Emergence::executeRequest()` to something low like 1 or 2
• Add $retries=1 parameter to Emergence::executeRequest() and have it return a re-invoktion of itselft with $retries- 1 if $retries > 0
• Evaluate different default values of the timeout and retries on a test instance with simulated parent failures

It should read /emergence/config.json and use the configured user/group

Mail handling has already been merged into Emergence/Skeleton but currently postfix has to be manually configured. A kernel service needs to be added to configure and launch postfix automatically.

Site::handleRequest should support special handlers in site-root:

• _index - used when the route matches its parent collection
• _notfound - used when a subpath was not found, the closest ancestor to the not found path is used
• _all - handles all requests within current collection and down even if a closer handler would match

#46 contains the first approach to integrating emergence
The new approach will be to:

• Continue installing emergence-kernel via npm install -g, so that both new machines can be initialized and old machines updated with the same command
• The kernel will handle installing habitat
  • maybe run the habitat curl|bash when the kernel starts if hab isn't found in the path?
  • maybe use some sort of npm hook to do that from a package.json option when the kernel is globally installed?
  • maybe repackage habitat as an npm package so it can just be a dependency for the kernel that npm handles installing to .bin
• The kernel will automate running hab commands
  • maybe just use a promise-based shell npm package to run hab commands
  • maybe repackage habitat as an npm package and expose async functions for each of the habitat commands we want that take node-ified options
• generate a single json document config for the entire host, consisting of the /emergence/config.json base document with a "sites" key added containing an object mapping each site handle under /emergence/sites/* to the contents of its site.json
• setup the config templates for each plan in https://github.com/JarvusInnovations/Emergence/tree/habitat/service-plans to generate from this shared mega-config tree
  • only the nginx config will loop through the "sites" list
  • we might want to remap services.plugins.myservice in legacy config.json to just services.myservice on load if plugins is detected the only key. Update the code that auto-generates new config.json to skip the plugins subkey going forward
• start the habitat supervisor if needed
• reimplement the start / stop functions for each of kernel-lib/services/* to load+start / stop via hab sup
• query hab sup status for current service status on startup and when requested instead of using pid files and maintaining a started/stopped state internally
• customize health check hooks for each service to do as complete a check as possible for being usable by emergence

This configuration will be needed:

ssl_session_cache shared:SSL:10m;

Workflow:

Run emergence-slave-setup on the slave instance
The script should

1. Creates an SSH account
2. Create an SSH key
3. Provide a script/command to run to add the slave on the master

Run emergence-master-setup on the slave instance

1. The script will instruct you to run emergence-slave-setup first and to cut and paste the script/command.
2. When you run the script/command it will connect via SSH to the slave instance and perform the following

If the kernel is configured to listen for web requests on an alt port, incorporate that into the links provided to sites in the kernel control UI

Example: "Getting started with an external client" in Help links to http://emr.ge/manual/getting_started_osx, which 404's.

Sort should be lexical by filename only, layer should be irrelevant. EventBus had similar issue recently for handlers that was resolved

Currently SSL vhosts only get configured with the primary_hostname under the assumption that that's the only one that will work with the cert. Wildcard certs might be used that that support many of the alt hostnames, so all hostnames should be written to the SSL vhost to ensure the correct site gets resolved for those hostnames

Update guide for installing nginx 1.6 and create an upgrade guide. Enable support for X-Accel-Buffering