Issue using environment variable for pipelines

[kago-dk]

Describe the bug
Environment variables used for pipelines are not processed correctly in 0.106.1, while they worked in 0.105.0. I have tried to put the value in double quotes and disable the confmap.strictlyTypedInput feature gate without success.

I a receiving an error instead.

When using confmap.strictlyTypedInput
failed to get config: cannot resolve the configuration: expanding ${env:OTEL_LOGS_RECEIVER}: retrieved value does not have unambiguous string representation: [filelog windowseventlog/application]

When using -confmap.strictlyTypedInput
failed to get config: cannot resolve the configuration: expanding ${env:OTEL_LOGS_RECEIVER}: expected convertable to string value type, got ["filelog" "windowseventlog/application"]([]interface {})

What version did you use?
0.106.1

What config did you use?

    logs:
      receivers: ${env:OTEL_LOGS_RECEIVER}
      processors: [memory_limiter,resourcedetection/system,resource/environment,transform,filter,batch]
      exporters: ${env:OTEL_LOGS_EXPORTER}

With [filelog,windowseventlog/application] as the environment value for OTEL_LOGS_RECEIVER as an example.

Environment
Windows 2022 Server

[kago-dk]

Tagging @mx-psi @TylerHelmuth per Slack request.

[mx-psi]

@kago-dk Could you share the exact value of your OTEL_LOGS_RECEIVER and OTEL_LOGS_EXPORTER environment variables? I would also be interested in your full configuration (censoring any sensitive details), or at least any other places where you use OTEL_LOGS_RECEIVER and OTEL_LOGS_EXPORTER.

[mx-psi]

Failed attempt at reproducing this (edited for clarity):

❯ cd $(mktemp -d)

❯ curl -L -o contrib-0.106.1.tar.gz <URL for contrib>
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
100 28.6M  100 28.6M    0     0  11.5M      0  0:00:02  0:00:02 --:--:-- 24.0M

❯ tar pfx contrib-0.106.1.tar.gz 

❯ cat config.yaml
receivers:
  otlp/1:
    protocols:
      grpc:
  hostmetrics:
    scrapers:
      load:

exporters:
  debug:

service:
  pipelines:
    logs:
      receivers: ${env:OTEL_LOGS_RECEIVER}
      exporters: [debug]

❯ OTEL_LOGS_RECEIVER=[otlp/1,hostmetrics] ./otelcol --config config.yaml 

<runs as usual>

[kago-dk]

The content respectively is [filelog,windowseventlog/application] and [splunk_hec,otlphttp/loki/abcd] for OTEL_LOGS_RECEIVER and OTEL_LOGS_EXPORTER. The values are set at Windows Environment Variables and not used elsewhere in the Otel Configuration.

I load the pipeline section into otel-config.yaml.

service:
  pipelines: ${file:${env:ProgramFiles}/otelcol/pipelines/pipelines.yaml}

If I paste the content from pipelines.yaml into otel-config.yaml (instead of using ${file:${env:ProgramFiles}/otelcol/pipelines/pipelines.yaml}) then the Collector is able to start up.

It sounds like the issue is resolved in 0.107.0 as part of a different ticket?

[mx-psi]

Yes, I think this is a duplicate of #10787. I am going to add a test case for this to verify it