* update vendor directory, add go.opencensus.io
* update imports
* oops
* s/opentracing/opencensus/ & remove prometheus / zipkin stuff & remove old stats
* the dep train rides again
* fix gin build
* deps from last guy
* start in on the agent metrics
* she builds
* remove tags for now, cardinality error is fussing. subscribe instead of register
* update to patched version of opencensus to proceed for now TODO switch to a release
* meh
fix imports
* println debug the bad boys
* lace it with the tags
* update deps again
* fix all inconsistent cardinality errors
* add our own logger
* fix init
* fix oom measure
* remove bugged removal code
* fix s3 measures
* fix prom handler nil
* Logs should support specifying region when using S3-compatible object store
* Use aws-sdk-go client for s3 backed logstore
* fixes vendor with aws-sdk-go dependencies
this rewrites the load balancer which was pointed out to be problematic here:
https://github.com/iron-io/functions/issues/570 with a test suite located at
test/fnlb-test-harness -- this test is now 'passing' in the sense that now
when running multiple functions nodes behind 1 load balancer every call goes
to the same functions node. yay. used a different consistent hashing algorithm
and also threw out all the fallback code (all the code, really).
this is basically an mvp and there is some work that needs to be done for
running multiple load balancers, allowing functions to run on more nodes as
load increases and some basic features like http/2 -- also needs some love to
be really robust; most of this is noted in TODOs in the file.
this does basic health checking configurable in the same way as aws elb. I
think we could probably do gossip but this works as a first cut. after N
health checks fail then requests start to go to a different node, meanwhile
all requests to that node will fail (need to fix). this continues to use
consistent hashing, which is great in that we don't need to store anything,
and we may be able to stick with the strategy in the future with some care.
