* Docs for Unix Domain Sockets
There are a number of cases where a user might wish to either 1)
expose a service through a Unix Domain Socket in the filesystem
('downstream') or 2) connect to an upstream service by a local unix
domain socket (upstream).
As of Consul (1.10-beta2) we've added new syntax and support to configure
the Envoy proxy to support this
To connect to a service via local Unix Domain Socket instead of a
port, add local_bind_socket_path and optionally local_bind_socket_mode
to the upstream config for a service:
upstreams = [
{
destination_name = "service-1"
local_bind_socket_path = "/tmp/socket_service_1"
local_bind_socket_mode = "0700"
...
}
...
]
This will cause Envoy to create a socket with the path and mode
provided, and connect that to service-1
The mode field is optional, and if omitted will use the default mode
for Envoy. This is not applicable for abstract sockets. See
https://www.envoyproxy.io/docs/envoy/latest/api-v3/config/core/v3/address.proto#envoy-v3-api-msg-config-core-v3-pipe
for details
NOTE: These options conflict the local_bind_socket_port and
local_bind_socket_address options. We can bind to an port or we can
bind to a socket, but not both.
To expose a service listening on a Unix Domain socket to the service
mesh use either the 'socket_path' field in the service definition or the
'local_service_socket_path' field in the proxy definition. These
fields are analogous to the 'port' and 'service_port' fields in their
respective locations.
services {
name = "service-2"
socket_path = "/tmp/socket_service_2"
...
}
OR
proxy {
local_service_socket_path = "/tmp/socket_service_2"
...
}
There is no mode field since the service is expected to create the
socket it is listening on, not the Envoy proxy.
Again, the socket_path and local_service_socket_path fields conflict
with address/port and local_service_address/local_service_port
configuration entries.
Set up a simple service mesh with dummy services:
socat -d UNIX-LISTEN:/tmp/downstream.sock,fork UNIX-CONNECT:/tmp/upstream.sock
socat -v tcp-l:4444,fork exec:/bin/cat
services {
name = "sock_forwarder"
id = "sock_forwarder.1"
socket_path = "/tmp/downstream.sock"
connect {
sidecar_service {
proxy {
upstreams = [
{
destination_name = "echo-service"
local_bind_socket_path = "/tmp/upstream.sock"
config {
passive_health_check {
interval = "10s"
max_failures = 42
}
}
}
]
}
}
}
}
services {
name = "echo-service"
port = 4444
connect = { sidecar_service {} }
Kind = "ingress-gateway"
Name = "ingress-service"
Listeners = [
{
Port = 8080
Protocol = "tcp"
Services = [
{
Name = "sock_forwarder"
}
]
}
]
consul agent -dev -enable-script-checks -config-dir=./consul.d
consul connect envoy -sidecar-for sock_forwarder.1
consul connect envoy -sidecar-for echo-service -admin-bind localhost:19001
consul config write ingress-gateway.hcl
consul connect envoy -gateway=ingress -register -service ingress-service -address '{{ GetInterfaceIP "eth0" }}:8888' -admin-bind localhost:19002
netcat 127.0.0.1 4444
netcat 127.0.0.1 8080
Signed-off-by: Mark Anderson <manderson@hashicorp.com>
* fixup Unix capitalization
Signed-off-by: Mark Anderson <manderson@hashicorp.com>
* Update website/content/docs/connect/registration/service-registration.mdx
Co-authored-by: Blake Covarrubias <blake@covarrubi.as>
* Provide examples in hcl and json
Signed-off-by: Mark Anderson <manderson@hashicorp.com>
* Apply suggestions from code review
Co-authored-by: Blake Covarrubias <blake@covarrubi.as>
* One more fixup for docs
Signed-off-by: Mark Anderson <manderson@hashicorp.com>
Co-authored-by: Blake Covarrubias <blake@covarrubi.as>
When info.Timeout is 0, it should have no timeout. Previously it was using a 0 duration timeout
which caused it to return without waiting.
This bug was masked by using a timeout in the tests. Removing the timeout caused the tests to fail.
This PR adds cluster members to the metrics API. The number of members per
segment are reported as well as the total number of members.
Tested by running a multi-node cluster locally and ensuring the numbers were
correct. Also added unit test coverage to add the new expected gauges to
existing test cases.
The intent of this struct was to prevent non-json output to stdout. With
the previous cleanup, this can now be done by simply changing the stdout
stream to io.Discard.
This is just one example of why passing around io.Writers for the
streams is better than the UI interface.
Previously this line was mixed up with logging, which made the output
quite ugly. Use the logger to output this message, instead of printing
directly to stdout.
This has the advantage that the message will be visible when json logs
are enabled.
Remove the leading whitespace on every log line. This was causing problems for
a customer because their logging system was interpretting the logs as a single
multi-line log.
The version args are static and passed in from the caller. Instead read
the static values in New.
The shutdownCh was never closed, so did nothing. Remove it as a field
and an arg.
We have seen test flakes caused by 'concurrent map read and map write', and the race detector
reports the problem as well (prevent us from running some tests with -race).
The root of the problem is the grpc expects resolvers to be registered at init time
before any requests are made, but we were using a separate resolver for each test.
This commit introduces a resolver registry. The registry is registered as the single
resolver for the consul scheme. Each test uses the Authority section of the target
(instead of the scheme) to identify the resolver that should be used for the test.
The scheme is used for lookup, which is why it can no longer be used as the unique
key.
This allows us to use a lock around the map of resolvers, preventing the data race.
* fix tests to use a dummy nodeName and not fail when hostname is not a valid nodeName
* remove conditional testing
* add test when node name is invalid
Normally the named pipe would buffer up to 64k, but in some cases when a
soft limit is reached, they will start only buffering up to 4k.
In either case, we should not deadlock.
This commit changes the pipe-bootstrap command to first buffer all of
stdin into the process, before trying to write it to the named pipe.
This allows the process memory to act as the buffer, instead of the
named pipe.
Also changed the order of operations in `makeBootstrapPipe`. The new
test added in this PR showed that simply buffering in the process memory
was not enough to fix the issue. We also need to ensure that the
`pipe-bootstrap` process is started before we try to write to its
stdin. Otherwise the write will still block.
Also set stdout/stderr on the subprocess, so that any errors are visible
to the user.
* debug: remove the CLI check for debug_enabled
The API allows collecting profiles even debug_enabled=false as long as
ACLs are enabled. Remove this check from the CLI so that users do not
need to set debug_enabled=true for no reason.
Also:
- fix the API client to return errors on non-200 status codes for debug
endpoints
- improve the failure messages when pprof data can not be collected
Co-Authored-By: Dhia Ayachi <dhia@hashicorp.com>
* remove parallel test runs
parallel runs create a race condition that fail the debug tests
* Add changelog
Co-authored-by: Daniel Nephin <dnephin@hashicorp.com>
Mark Anderson