feat: socketstat input plugin

[sajoupa]

Required for all PRs:

• Signed CLA.
• Associated README.md updated.
• Has appropriate unit tests.

Hi,

The ss command (package iproute2) gives very useful metrics about sockets on Linux machines, this is a plugin to expose them with telegraf.

Thanks.

[danielnelson]

@sajoupa Thanks for the pull request. I think there may be two high level concerns with this pr:

• Since a new series is made for every state+src+dst, this output format will be very high cardinality, you will probably need to pair this with a short retention policy to avoid overloading your database.
• Shelling out to ss will be somewhat expensive, perhaps we can bypass this and directly retrieve the data from /proc or another method.

[sajoupa]

Hi @danielnelson , you're totally right about data cardinality, I added a warning to the README.
I'm quite reluctant to parse /proc/net/ however. We would duplicate (and then maintain) logic which is already implemented by ss. I don't think that it would be worth it, given the small cost of shelling out to ss.

I have also updated the code to make the same changes which had been requested for the ipset input plugin (use a timeout, buffio scanner, ... ref #3346)

[danielnelson]

I'm not sure if I will want to merge this in, in the meantime we can leave the pr open for anyone who would like to use it.

[sajoupa]

@danielnelson which of the two reasons you first mentioned would block this ?
I can replace shelling out to ss by parsing /proc (I believe it's better to shell out, but both solutions work), however, there's no reducing data cardinality if we want to get ss data.
My intent is to have ss exported data sent to graylog. According to https://github.com/Graylog2/graylog-plugin-netflow and https://www.graylog.org/post/snmp-netflow-plug-ins-extend-graylog-network-infrastructure-support, graylog can handle the cardinality of network packets metadata.
I just find it a lot more convenient to get data from telegraf agents than network equipments. (also, activating netflow on network equipments won't help for cloud VMs).
Thanks.

[danielnelson]

Shelling out to ss is the bigger issue, we can add warnings to the documentation about cardinality and how to deal with it.

[sajoupa]

@danielnelson Hi, I looked further into it and think shelling out is the only option.
ss uses the tcp_diag kernel module to get its data much faster than parsing /proc/net (https://www.cyberciti.biz/files/ss.html). Also, /proc/net won't provide counters such as bytes_acked, received, etc.
And after looking at ss's 4900+ lines of code, I'm not sure we want to recreate them into telegraf ;)

[srebhan]

Thanks for your patience @sajoupa! There are a few additional improvements I'd like to see going in. Some are just a matter of taste, so please speak-up if you don't like them.

[telegraf-tiger]

☺️ This pull request doesn't significantly change the Telegraf binary size (less than 1%)

📦 Looks like new artifacts were built from this PR.

Expand this list to get them here ! 🐯

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_i386.zip
armel.deb	armv6hl.rpm	freebsd_amd64.tar.gz
armhf.deb	i386.rpm	freebsd_armv7.tar.gz
i386.deb	ppc64le.rpm	freebsd_i386.tar.gz
mips.deb	s390x.rpm	linux_amd64.tar.gz
mipsel.deb	x86_64.rpm	linux_arm64.tar.gz
ppc64el.deb		linux_armel.tar.gz
s390x.deb		linux_armhf.tar.gz
		linux_i386.tar.gz
		linux_mips.tar.gz
		linux_mipsel.tar.gz
		linux_ppc64le.tar.gz
		linux_s390x.tar.gz
		static_linux_amd64.tar.gz

[srebhan]

Awesome! Thanks for driving this home @sajoupa! LGTM.

[sajoupa]

Awesome! Thanks for driving this home @sajoupa! LGTM.

Cheers, thanks for the patient reviews.