This PR switches the Nomad repository from using govendor to Go modules
for managing dependencies. Aspects of the Nomad workflow remain pretty
much the same. The usual Makefile targets should continue to work as
they always did. The API submodule simply defers to the parent Nomad
version on the repository, keeping the semantics of API versioning that
currently exists.
We have been using fatih/hclfmt which is long abandoned. Instead, switch
to HashiCorp's own hclfmt implementation. There are some trivial changes in
behavior around whitespace.
Use v1.1.5 of go-msgpack/codec/codecgen, so go-msgpack codecgen matches
the library version.
We branched off earlier to pick up
f51b518921
, but apparently that's not needed as we could customize the package via
`-c` argument.
Examples for HTTP based task-group service healthchecks are
covered by the `countdash` demo, but gRPC checks currently
have no runnable examples.
This PR adds a trivial gRPC enabled application that provides
a Service implementing the standard gRPC healthcheck interface.
Running `make dev` runs `hclfmt`, but this isn't checked as part of
CI. That makes it possible to merge un-formatted HCL and Nomad
jobspecs that later will make for dirty git staging areas when
developers pull master.
This changeset adds HCL linting to the `make check` target.
Use go mod for github.com/hashicorp/go-bindata/go-bindata and
github.com/elazarl/go-bindata-assetfs/go-bindata-assetfs but use
`@master` to pull the latest master. These packages don't have release
tags so `@master` worksaround it.
Adding `-trimpath` to builds removes the local working directory from
the goroutine stack traces, which makes our builds more reproducible
and doesn't leak information about our local development workstations
or CI environment.
This allows using https download and go mod cache proxies, over using
git and downloading entire dependencies git history, hopefully,
resulting into a faster installation process.
You'd think since golangci-lint embeds misspell we could use that,
but it fails to run if it finds no Go source files, which is the
case in our website/ directory that we want to check.
gometalinter has been deprecated, with golangci-lint as its spiritual
and recommended successor. Here we switch to using it with an equivalent
configuration, albeit with newer versions of some linters.
To maintain compatibility with existing settings, we have a couple of
things disabled here, specifically:
- tests
We have a lot of unused code in our tests that choke deadcode.
We should attempt to clean these up soon so that we can lint our
testcode.
- govet.check-shadowing = false
This breaks on redefining `err` which we do all over the nomad
codebase.
This is an attempt to ease dependency management for external driver
plugins, by avoiding requiring them to compile ugorji/go generated
files. Plugin developers reported some pain with the brittleness of
ugorji/go dependency in particular, specially when using go mod, the
default go mod manager in golang 1.13.
Context
--------
Nomad uses msgpack to persist and serialize internal structs, using
ugorji/go library. As an optimization, we use ugorji/go code generation
to speedup process and aovid the relection-based slow path.
We commit these generated files in repository when we cut and tag the
release to ease reproducability and debugging old releases. Thus,
downstream projects that depend on release tag, indirectly depends on
ugorji/go generated code.
Sadly, the generated code is brittle and specific to the version of
ugorji/go being used. When go mod picks another version of ugorji/go
then nomad (go mod by default uses release according to semver),
downstream projects face compilation errors.
Interestingly, downstream projects don't commonly serialize nomad
internal structs. Drivers and device plugins use grpc instead of
msgpack for the most part. In the few cases where they use msgpag (e.g.
decoding task config), they do without codegen path as they run on
driver specific structs not the nomad internal structs. Also, the
ugorji/go serialization through reflection is generally backward
compatible (mod some ugorji/go regression bugs that get introduced every
now and then :( ).
Proposal
---------
The proposal here is to keep committing ugorji/go codec generated files
for releases but to use a go tag for them.
All nomad development through the makefile, including releasing, CI and
dev flow, has the tag enabled.
Downstream plugin projects, by default, will skip these files and life
proceed as normal for them.
The downside is that nomad developers who use generated code but avoid
using make must start passing additional go tag argument. Though this
is not a blessed configuration.
We currently use an container image for `test-devices` job only; while
all other jobs use machine executor.
This allows us to switch golang and protoc verions easily without
manually managing Docker images (which requires building them manually
on a dev machines, etc). All that while, we install dependencies on
every build in all other jobs..
`test-devices` now is one of the fastest jobs and isn't a constraint or
a bottleneck, so increasing its overhead by few seconds doesn't hurt the
overall developer iteration.
If we split tests effectively later, we can revisit.
Allow honoring `GO_TAGS` environment variable if set. Currently, users
must set variable as a makefile argument e.g. `make GO_TAGS=ui dev`, and
this allows us to use env-var syntax (e.g. `GO_TAGS=ui make dev`) and
make it convenient to set GO_TAGS globally.
This commit provides an initial migration of general testing CI
infrastructure to CircleCI.
It uses CircleCI 2.1 paramereterised jobs to provide two base
configurations: a vm based `test-machine`, and docker based
`test-container`.
Jobs that require root, docker, or other similar features require the
machine based jobs, but others should be ran using the `test-container` package
as they are both cheaper and faster to run.
Our build scripts pass `$(GO_TAGS)` to `-tags` go compile flags, except
for `make dev`, where `$(NOMAD_UI_TAG)` is used. This change ensures
`make dev` is inline with the rest of makefile targets.
I use the flag primarily to enable the nomad ui using the committed
compiled assets without regenerating them, as I find using stale ui
satisfactory most of the time.
`make check` runs very intensive linters that slow and seem to behave
differently on different machines.
Linting is still a part of our CI and we shouldn't be cutting a release
when CI isn't green anyway.
Mahmood Ali