Skip vendor-specific statistics that lead to duplicate metric labels

[lahwaacz]

fixes #3

[baryluk]

Thanks. This fixed issues on Samsung Exos X16 for me.

And I think it is a good fix. There is no point of collecting the metrics for these statistics, which just would consume space, and never be used.

Lets merge it please. It can be always improved in the future to make sense of these metrics other way.

[kfox1111]

Nice. We just ordered some Samsung Exos X16's. So good to know that this is a problem / there is a fix for it. Thanks!

[vagifzeynalov]

doesn't look right to me.
"avoid duplicates" doesn't mean "ignore all of them"

[vagifzeynalov]

#32 - alternative solution

[lahwaacz]

Thanks. This fixed issues on Samsung Exos X16 for me.

Just for the record, it's Seagate, not Samsung.

"avoid duplicates" doesn't mean "ignore all of them"

Do you have some use case for the Vendor Specific Statistics? As @baryluk said, it does not make sense to collect metrics that we can't make sense of.

[vagifzeynalov]

No, I personally don't have any use cases for those values.
But removing them also doesn't look good to me. Who knows how they could be used? 😁
As I said in my PR, I'll try to look at them closer...

[baryluk]

But removing them also doesn't look good to me. Who knows how they could be used? grin

For wasting storage, memory, CPU cycles and energy....

So, they should be removed.

Good metrics, should be actually useful and designed to be so. I would say all of the smart statistics (not attributes and other stuff) should be behind the command line flag, and be disable by default.

[vagifzeynalov]

For wasting storage, memory, CPU cycles and energy....

Nowadays hardware is cheap, so it is not an argument to save few megabytes.

Good metrics, should be actually useful and designed to be so. I would say all of the smart statistics (not attributes and other stuff) should be behind the command line flag, and be disable by default.

One thing to check something from the command line, another thing to have historical data and check it only if/when needed.
Perhaps there could be a configuration file where all kinds of metrics could turned ON/OFF.

[retrry]

Can we get a resolution on this? It is better to drop these statistics, than to try to collect them and fail. smartctl_exporter currently is unusable on some of the drives because of this issue.

[Stogas]

@vagifzeynalov Would you consider adding an "ignore duplicates" runtime flag?

You could leave it to default to the current behavior, while allowing affected users who don't mind losing vendor-specific data to easily drop it without recompiling their own exporter

[onedr0p]

@SuperQ any chance we can see some effort on getting this merged?

[k0ste]

@lahwaacz can you rebase this PR? This should trigger CI
Thanks!

[lahwaacz]

@k0ste rebased

[SuperQ]

This needs a DCO sign-off. You can use git commit -s --amend to add it.

[k0ste]

@lahwaacz https://app.circleci.com/pipelines/github/prometheus-community/smartctl_exporter/92/workflows/454640dc-30ac-4e42-8d3d-843f5f7f4c5e/jobs/264