This PR uses the checksum of the check for which a dynamic exposed
port is being generated (instead of a UUID prefix) so that the
generated port label is deterministic.
This fixes 2 bugs:
- 'job plan' output is now idempotent for jobs making use of injected ports
- tasks will no longer be destructively updated when jobs making use of
injected ports are re-run without changing any user specified part of
job config.
Closes: https://github.com/hashicorp/nomad/issues/10099
The expose handler hook must handle if the submitted job is invalid. Without this validation, the rpc handler panics on invalid input.
Co-authored-by: Tim Gross <tgross@hashicorp.com>
This PR enables jobs configured with a custom sidecar_task to make
use of the `service.expose` feature for creating checks on services
in the service mesh. Before we would check that sidecar_task had not
been set (indicating that something other than envoy may be in use,
which would not support envoy's expose feature). However Consul has
not added support for anything other than envoy and probably never
will, so having the restriction in place seems like an unnecessary
hindrance. If Consul ever does support something other than Envoy,
they will likely find a way to provide the expose feature anyway.
Fixes#9854
In #7800, Nomad would automatically generate a port label for service
checks making use of the expose feature, if the port was not already
set. This change assumed the group network would be correctly defined
(as is checked in a validation hook later). If the group network was
not definied, a panic would occur on job submisssion. This change
re-uses the group network validation helper to make sure the network
is correctly definied before adding ports to it.
Fixes#8875
Part of #6120
Building on the support for enabling connect proxy paths in #7323, this change
adds the ability to configure the 'service.check.expose' flag on group-level
service check definitions for services that are connect-enabled. This is a slight
deviation from the "magic" that Consul provides. With Consul, the 'expose' flag
exists on the connect.proxy stanza, which will then auto-generate expose paths
for every HTTP and gRPC service check associated with that connect-enabled
service.
A first attempt at providing similar magic for Nomad's Consul Connect integration
followed that pattern exactly, as seen in #7396. However, on reviewing the PR
we realized having the `expose` flag on the proxy stanza inseperably ties together
the automatic path generation with every HTTP/gRPC defined on the service. This
makes sense in Consul's context, because a service definition is reasonably
associated with a single "task". With Nomad's group level service definitions
however, there is a reasonable expectation that a service definition is more
abstractly representative of multiple services within the task group. In this
case, one would want to define checks of that service which concretely make HTTP
or gRPC requests to different underlying tasks. Such a model is not possible
with the course `proxy.expose` flag.
Instead, we now have the flag made available within the check definitions themselves.
By making the expose feature resolute to each check, it is possible to have
some HTTP/gRPC checks which make use of the envoy exposed paths, as well as
some HTTP/gRPC checks which make use of some orthongonal port-mapping to do
checks on some other task (or even some other bound port of the same task)
within the task group.
Given this example,
group "server-group" {
network {
mode = "bridge"
port "forchecks" {
to = -1
}
}
service {
name = "myserver"
port = 2000
connect {
sidecar_service {
}
}
check {
name = "mycheck-myserver"
type = "http"
port = "forchecks"
interval = "3s"
timeout = "2s"
method = "GET"
path = "/classic/responder/health"
expose = true
}
}
}
Nomad will automatically inject (via job endpoint mutator) the
extrapolated expose path configuration, i.e.
expose {
path {
path = "/classic/responder/health"
protocol = "http"
local_path_port = 2000
listener_port = "forchecks"
}
}
Documentation is coming in #7440 (needs updating, doing next)
Modifications to the `countdash` examples in https://github.com/hashicorp/demo-consul-101/pull/6
which will make the examples in the documentation actually runnable.
Will add some e2e tests based on the above when it becomes available.
hashicorp-copywrite[bot]