prometheus: use older node_exporter

[tbg]

v1.3.1, the most up to date released version, has a bug that inflates
the bytes written by ~8x for NVMe drives (which in particular includes
the default drives for our GCE roachprod machines). Fundamentally this
is caused by the fact that these devices use a 4K sector size whereas
the kernel will always report based on a 512B sector size.

This took us a while to figure out, and to avoid repeating this exercise
periodically, downgrade node_exporter to 1.2.2, which pre-dates a
refactor that introduces the regression.

See: prometheus/node_exporter#2310

Release note: None

[cockroach-teamcity]

This change is 

[nicktrav]

Thank you!

Reviewed 1 of 1 files at r1, all commit messages.
Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @tbg)

[tbg]

bors r=nicktrav

[srosenberg]

v1.3.1, the most up to date released version, has a bug that inflates the bytes written by ~8x for NVMe drives (which in particular includes the default drives for our GCE roachprod machines). Fundamentally this is caused by the fact that these devices use a 4K sector size whereas the kernel will always report based on a 512B sector size.

This took us a while to figure out, and to avoid repeating this exercise periodically, downgrade node_exporter to 1.2.2, which pre-dates a refactor that introduces the regression.

Yep, I remember pulling some hairs out when I ran into the same issue with fio [1]. I'll also update the packer base image to 1.2.2 so that (soon) teamcity agents will be running it too.

[1] axboe/fio#1325 (comment)

[craig]

Build succeeded:

• GitHub CI (Cockroach)