Call a specific job scrape via a query string

[tchabit]

Hello,

I'm new to prometheus and trying to have a different scrapping interval for collectors, without having to configure two sql_exporter services.

Despite using the exemple here : #37 (comment) , I did not find a way for prometheus to call one specific 'job' defined in sql_exporter.yml

jobs:
  - job_name: test1
    collectors: [ports]
    static_configs:
      - targets:
          data1: 'mysql://test:test@localhost:3306/test'

  - job_name: test2
    collectors: [devices]
    static_configs:
      - targets:
          data2:  'mysql://test:test@localhost:3306/test'

With this configuration it add an instance -target db name) and job label but I dont find how to call only one "sql_exporter job" from prometheus or through curl as I dont find a way to make params works.

Am I missing something ?

[burningalchemist]

Hey @tchabit, unfortunately it won't work this way.

Scraping for this particular exporter is controlled by Prometheus, so scrapes are performed when Prometheus sends a request to the exposed endpoint. For the time being there is no way to call a specific job - all of them will be processed.

It's an interesting use case, and can be considered as a feature, but needs some investigation before the implementation to avoid potential regressions. So I can't promise any ETA.

For now, you might try to configure two separate scrape jobs in Prometheus with different intervals using the same endpoint, but it's rather a workaround than a solution. Or two separate instances, which is not preferable in your case.

Kind regards,
Sergei

[tchabit]

Hello @burningalchemist

Thank you for the prompt answer and yes in my opinion the possibility to use prometheus "params" could be a great improvement.

Regarding the proposed workaround, I dont get how it might work if the endpoint is the same, I'll juste have the same collectors scrapped with different frequency isnt ?
For the time being I've used another instance (separate config file to select collectors + different port) but its quite inelegant.

Have a nice day

[burningalchemist]

Hey @tchabit,

Yeah, exactly, same endpoint, different intervals.

The footprint of exporter is small enough (it's possible to make it very small, if you compile your version with the only required drivers), so it shouldn't be a big issue to run multiple instances.

I'll see what I can do about it and will mention you once I get some results. Thanks for the suggestion! 👍

[RABBITERMAN]

hi dear,i have same problem.i want to use a one sqlexporter for to collector sepratly from two prometus servers.
is it possible?

[Emptyfruit]

Regarding initial question, can't this be achieved defining different collectors and setting different min_interval values on each of them?
So prom will scrape every 15s but the real requests will only execute every min_interval. Or am i misunderstanding how this works?

[burningalchemist]

@Emptyfruit, apologies for the late reply. Yes, indeed, this should work just fine. While it won't solve the issue when a user wants to use the same collector against multiple instance sets and call them separately, it overall gives the desired result. Thank you! 👍

@RABBITERMAN @tchabit You might want to try the option above if it's still relevant to you. 🙂 As for querying specific jobs, there's still no ETA so far. I've sketched some ideas, but haven't found time to test them yet.

[burningalchemist]

@tchabit just for your information: I've finally found some time to prototype this feature. The prototype works fine, but we'll need to look deeper into the general integration, naming, regression testing, etc. So it might be introduced in v0.14 release.

[redzioch]

+1

Similar use case - same target, different intervals.

[tchabit]

Sorry for the lack of response. Regarding @Emptyfruit suggestion it'd work but consume TSDB storage, something problematic as we keep those metrics for months.
@burningalchemist Great, I'm looking forward those improvements, let me know if you want me to test them out :)

[barywhyte]

Hello @burningalchemist

I do not know weather to raise this as an issue but I saw in release note for v0.12.2 that the context deadline issue was addressed.

For this is what I get whenever I attempt to collect metrics from localhost:9399/metrics.

No metrics gathered, [from Gatherer #1] context deadline exceeded

My datasource line looks like:

data_source_name: 'sqlserver://prom_usr:#!$deasgfhf@10.0.41.101:1433'

My database is MS SQL

I have changed scrape_timeout to about 60s and the error became complicated:

No metrics gathered, [from Gatherer #1] lookup prom_usr on 10.0.0.71:53: read udp 10.0.41.101:55098->10.0.0.71:53: i/o timeout

Both sql_exporter and MS SQL are running on the same windows host server 2016. I do not know what I am doing wrong here