feat(inputs.kernel): Add Pressure Stall Information

[iBug]

Summary

Implemented an input plugin for Pressure Stall Information (Linux v4.20+).

Part of the code was derived from gridscale/linux-psi-telegraf-plugin, which is available under the MIT license.

I will be after this PR as long as I am around.

Checklist

• No AI generated code was used in this PR

Related issues

Resolves #6760

[srebhan]

Thanks for your contribution @iBug! Before giving this a full review I would like to ask why you do not implement this feature as part of the inputs.kernel plugin? You could easily add this as a new collect ID and make this explorable to all users of kernel which is probably the target audience anyway!?!?

Furthermore, I suggest to merge the pressure and pressureTotal metrics to something like

pressure,resource=cpu some_avg10=1.53,some_avg60=1.87,some_avg300=1.73,some_total=1088168194,full_avg10=0,full_avg60=0,full_avg300=0,full_total=0 1700000000000000000
pressure,resource=memory some_avg10=0,some_avg60=0,some_avg300=0,some_total=3463792,full_avg10=0,full_avg60=0,full_avg300=0,full_total=1429641 1700000000000000000
pressure,resource=io some_avg10=0,some_avg60=0,some_avg300=0,some_total=68568296,full_avg10=0,full_avg60=0,full_avg300=0,full_total=54982338 1700000000000000000

What do you think @iBug?

[iBug]

implement this feature as part of the inputs.kernel plugin

Good suggestion. My reasons for not doing this now:

• I didn't realize this was possible when I started this PR.
• Maybe this could share some reasoning behind having separate plugins for inputs.system and inputs.swap, the latter of which completely relies on code from the former anyway.
• github.com/prometheus/procfs (which Telegraf is already using) provides a convenient interface for retrieving these data so I don't have to write yet another parser (albeit easy). I don't see a straightforward way to integrate the current implementation into inputs.kernel.

May I ask what should I do now? Start migrating this entire code to inputs.kernel, wait for more reviews / suggestions / discussions, or prepare a separate PR?

merge the pressure and pressureTotal metrics

• Merging pressure and pressureTotal makes total sense.
• Merging some and full tags: Hesitant for two reasons:
  • The kernel provides them on separate lines with the same structure, so logically keeping them apart seems better. This also works better for InfluxDB GROUP BY queries.
  • Specifically for resource=cpu, there is no type=full. Should I just omit the full_* values for resource=cpu?

---

Just revisited the code:

Technical difficulty in merging pressure and pressureTotal: The total field is a counter type while avg* fields are gauge types. I can add them all onto the same pressure series (in InfluxDB parlance) but I still need to AddCounter and AddGauge separately.

[srebhan]

@iBug let me discuss this with the team...

[srebhan]

@iBug we discussed this PR yesterday and can live with the current metric format to keep the Gauge vs. Counter properties. However, we do also agree on merging this plugin into inputs.kernel as this is a natural place for this information. I don't think this is too hard as you basically need to move the contents of psi_linux.go there and guard the execution of the function with optCollect containing the feature flag (e.g. "psi")... What do you think?

I can assist you in the merge if you want and permit pushes to your PR branch in your repository...

[iBug]

I just moved all the code into inputs.kernel and I think this PR is good to go.

permit pushes to your PR branch

I forked the repository to an organization named iBug-forks, and it seems like "allow edits from maintainers" is only for user-owned repos. Meanwhile, GitHub doesn't allow changing the incoming branch once a PR is created, so I'm afraid I'll have to commit & push all the code myself.

Fortunately I'll be able to spare some minutes daily in the upcoming days, so you can just comment and I'll handle the code. Alternatively I can close this PR, re-fork and re-submit another PR so you can push.

[srebhan]

Thanks for merging this to the kernel plugin @iBug! Two comments from my side...

[srebhan]

Very nice @iBug, it's really a joy to work with you! I would like to convert the test to use testutil.RequireMetricsEqual as this prints a nice diff on fail and allows to specify the expected metrics as such. The other two comments are minor...

We are almost there... :-)

[telegraf-tiger]

Download PR build artifacts for linux_amd64.tar.gz, darwin_arm64.tar.gz, and windows_amd64.zip.
Downloads for additional architectures and packages are available below.

🥳 This pull request decreases the Telegraf binary size by -3.74 % for linux amd64 (new size: 211.6 MB, nightly size 219.9 MB)

📦 Click here to get additional PR build artifacts

Artifact URLs

DEB	RPM	TAR GZ	ZIP
amd64.deb	aarch64.rpm	darwin_amd64.tar.gz	windows_amd64.zip
arm64.deb	armel.rpm	darwin_arm64.tar.gz	windows_arm64.zip
armel.deb	armv6hl.rpm	freebsd_amd64.tar.gz	windows_i386.zip
armhf.deb	i386.rpm	freebsd_armv7.tar.gz
i386.deb	ppc64le.rpm	freebsd_i386.tar.gz
mips.deb	riscv64.rpm	linux_amd64.tar.gz
mipsel.deb	s390x.rpm	linux_arm64.tar.gz
ppc64el.deb	x86_64.rpm	linux_armel.tar.gz
riscv64.deb		linux_armhf.tar.gz
s390x.deb		linux_i386.tar.gz
		linux_mips.tar.gz
		linux_mipsel.tar.gz
		linux_ppc64le.tar.gz
		linux_riscv64.tar.gz
		linux_s390x.tar.gz

[powersj]

Thank you very much for the back and forth on this PR!

[srebhan]

Thanks for all you work @iBug! It was fun to work with you! Looking forward to your next PR. :-)

[ajfriesen]

Thank you very much for integrating the psi plugin we created.

Very much appreciated 🙏