fission-bundle is a single binary that can be run as one or more of
the router, controller and poolmgr. It's built into one docker image
which can then be run with different commands.
Router uses poolmgr to specialize pods when necessary.
Router uses controller to get the list of triggers to listen for. For
now this integration is pretty crappy -- we just poll the controller
every few seconds and cache the result. The right way would be to
have some sort of watch API on the controller and use that. Or maybe
share access to etcd directly.
Switch to client-go package instead of pulling the from kubernetes.
Use a versioned client with sensible compatiblity.
This change breaks 'go get'. For now you have to manually checkout
the 'release-1.4' branch of the client-go package after 'go get'
fetches it. TODO: use one of the build tools to fix this.
GenericPool is a pool of generic containers for an environment.
GenericPoolManager keeps track of all GenericPools, creating them
on-demand.
The pool manager API is simply a "lookup" for the service URL of a
function. If one exists it is returned immediately; otherwise, a
generic pool is created, and then a pod is specialized from that pool.
poolmgr is designed to run from within the cluster, since it connects
to pod IP addresses directly.
This is a first cut with many pieces missing.
TODO:
* Use versioned kubernetes clients instead of the unversioned one
* Unit tests for GenericPoolMgr; improve unit test for GenericPool;
test for API.
* Handle cases where a service exists but pod backing it has failed.
* On start up, use existing deployments/pods/services if they exist;
in other words don't orphan resources on restart.
* Kill idle resources (services, pods, even generic pools)
* Autoscale generic pool (for example, by watching num ready pods)
Fetcher is a tool to download functions from the controller. It is
meant to be run in a container in the same pod as the function run
container, with a shared volume.
Environments for functions. For now this is just the image URL for the
function run container. In the future, Environments will also have
build container image URLs, and perhaps parameters to those
containers.
Create/read/update/delete/list. Versioning will be added later. For
now, only the latest update is stored.
This implemenents create/read/update/delete/list for http triggers.
This isn't quite the final implementation: you can't look up old
versions by UUID since it only stores the latest update. I'll get to
that later.
FunctionStore uses fileStore for function code and resourceStore for
the Function's metadata. Environment- and HTTPTrigger- Store are just
thin wrappers on resourceStore.
There is probably a better way to organize this code, maybe with
reflection.
- A generic Resource interface, which is serializable and has
a key() method to get an identifying key.
- A resourceStore, which has create, read, update, delete and list.
This wraps etcd key operations and serialization/deserialization in
a convenient service. Serialization is provided to the
resourceStore as a 'serialization' interface. This can be JSON or
anything else.
- The resourceStore also has file operations -- this has read, write
and delete operations. The general idea is to store metadata in
etcd and data on the fileStore (which could also be an object store,
we don't really use any file-specific properties). Files are also
versioned -- each writeFile writes to a new UUID and that UUID is
stored in an etcd in-order list. readFile can retrieve an arbitrary
version by UUID, or just get the latest contents.
This is currently:
1. A list of http triggers
2. A method to generate a mux router from that list
This will be in the future:
1. A way to keep the set of triggers up to date, by watching the controller API
2. A way to update the router when the trigger set changes
The function handler will be the handler for each http trigger. It
will look up the function service map to find a service for the
function (and call the pool manager to create a new service if one
doesn't already exist, though this isn't implemented yet). It'll then
proxy the request to the service.
Concurrency-safe mapping from a function's (name, uid) identifier to a
URL. This will be used to look up the service URL associated with a
given function.
Mutable mux test occasionally panics because the server is shut down
before the client goroutine. Fix this by shutting down the client
goroutines before shutting down the server.
We want to be able to update routes without restarting the http
server. mutableRouter is a very thin wrapper around
github.com/gorilla/mux that enables safely switching to a new mux
router without pausing requests.
First cut of a function run container for NodeJS. So far, just the
simplest stuff: loads a user function and routes requests to it. The
user function is assumed to be in a file at a provided location. The
fission runtime is expected to place the file there securely (exactly
how we'll do that is TBD).
Doesn't handle path template params, query strings, etc. Will also
need to be extended later with some sort of hook for graceful
shutdown.
Soam Vasani