systemd Vector Service Configuration with Environment Variables

[tbenade]

Hi,

I was hoping to use some templated configuration only swapping out the value of environment variables in /etc/default/vector on Linux EC2 instances running vector as a systemd service. We already used /etc/default/vector to configure some VECTOR_* values.

Deployed a quick POC and ran great. Then noticed in the validate output that we have a deprecation warning.

Loaded with warnings ["/etc/vector"]
------------------------------------
~ Unknown environment variable in config. This is DEPRECATED and will become an error in future versions. name = "SPLUNK_INDEX"

Two questions then. Maybe this guidance may be misleading for anyone comes across it. https://github.com/vectordotdev/vector/blob/master/distribution/systemd/vector.default#L2

Next question is what is the appropriate mechanism for despatching environment variables for a vector systemd service if not /etc/default/vector please. Apologies if this is purely a linux question, systemd is not in my regular skillset.

Thanks

[jszwedko]

/etc/environment/vector will automatically be loaded by the default SystemD unit:

vector/distribution/systemd/vector.service

Line 16 in 58a4a2e

EnvironmentFile=-/etc/default/vector

. Do you have a line like SPLUNK_INDEX=some_index in it?

[tbenade]

Correct the file is loaded and variables in the file are correctly loaded. My question is why the deprecation notice for "unknown" (custom variables) when running validate. Is the EnvironmentFile not the best mechanism for defining systemd environment variables?

[jszwedko]

Correct, undefined variables in Vector configuration will become a hard error in the future. You can provide a default to preserve the current behavior with bash-like syntax. For example: ${SPLUNK_INDEX-}.

[tbenade]

Thanks for that. I'm building new so not looking to introduce a regression. Where else does one define environment variables for vector aggregator running as a service if we can't use an EnvironmentFile. Thanks

[tbenade]

Might try answer my own question in an attempt to further my own learning.

I think I need to create an overrides service unit file which will point to an additional EnvironmentFile where I will be able to set my environment variables without upsetting validate?

[tbenade]

Yikes, I now realise I totally misunderstood. The error is not being generated by the loading of the systemd unit file but the parsing of my vector configuration 😵‍💫 . Sorry for being so confusing. In the end my issue was with the fact in my vector configuration in this case the Splunk HEC configuration I was referencing an "unknown" environment variable.

Thanks for the guiding me back to sanity.

[tbenade]

Documentation clearly explains Jesse's suggestion, https://vector.dev/docs/reference/configuration/#default-values