prometheus / prometheus/client_python

`InfoMetricFamily`’s `add_metric()` parameters are strangely named and should provide default values

Aperta
#1,049 1 commento 1 reazione 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Lingua principale
Python
Stelle
4.4k
Fork
876
Merge medio
8g 4h
PR unite (30g)
1

Descrizione

Hey.

I was looking into InfoMetricFamily’s add_metric() and I either I just don't get it or the API is a bit strange.

With an Info object I'd have done something like this:

m = prometheus_client.Info("logical_drive", "foobar", ("controller_name",
                                                       "array_name",
                                                       "logical_drive_name",
                                                       "caching",
                                                       "device",
                                                       "raid_level",
                                                       "logical_drive_label",
                                                       "multidomain_status",
                                                       "parity_initialization_status",
                                                       "status"
                                                      )
                      )

and then set it like:

m.labels(controller_name=controller_name, array_name=array_name, logical_drive_name=logical_drive_name,
         **{n: logical_drive_properties.get(n, "")  for n in (
                                                              "device",
                                                              "raid_level",
                                                              "logical_drive_label",
                                                              "multidomain_status",
                                                              "parity_initialization_status",
                                                              "status"
                                                             )
           },
         caching=bool_or_none_to_label_value( logical_drive_properties.get("caching") )
        )

(here an example where I set some of the properties directly, and some via dict unpacking)

  • It would fail if I forgot a label.
  • I could use label names as attribute names, like controller_name= rather than "controller_name"=... which is however not super important

With InfoMetricFamily seem quite a bit different:

  • add_metric() now has the parameters labels and value.
  • As far as I understand the code, labels are actually not labels, but values for these, namely the ones set with labels in the constructor of InfoMetricFamily.
  • It's no longer possible to give labels as dict, one really needs to give them in the right order as sequence. Why does Info allow that but not this?
  • AFAICS, there is no check if exactly those labels are set, that were given in the constructor. Why over at Info but not here? What sense does it then even make to set the labels in the constructor?
  • value is misleading in so many ways. It's not the value of the metric (that is 1) and even if it relates to the labels being the “value”, then it should be values.
  • It might have perhaps even been better to swap the two names... or rater use some completely different names.

Anyway... guess this can't be changed now without breaking the ABI.

However, as far as I understand the code:
https://github.com/prometheus/client_python/blob/09a5ae30602a7a81f6174dae4ba08b93ee7feed2/prometheus_client/metrics_core.py#L372

the idea is one can give labels and/or value and both are merged in a final dict of label/value pairs.... but then it would be nice if labels and value were not required.

Why not simply give them default values () respectively {}?

I could provide a patch if nothing speaks against that particular change. But the above points are IMO still problematic. Especially also that the behaviour is considerable different from Info, which is not really obvious.

Cheers,
Chris.

Guida per i contributori

Apri la guida per i contributori

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Direzione di ricerca

Inizia con l’implementazione collegata di InfoMetricFamily.add_metric() in prometheus_client/metrics_core.py intorno alla riga 372, quindi confronta la gestione dei relativi parametri con Info. Conferma i valori predefiniti previsti e le implicazioni di compatibilità per labels e value; il lavoro è completato quando una modifica dell’API concordata è implementata senza interrompere i chiamanti esistenti e il suo comportamento è coperto dai test.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Valutazione

Stack tecnologico
python
Ambito
observability-sre
Tipo di issue
Funzionalità
Difficoltà
4/5
Tempo stimato
3-5 giorni
Stato di attività
Ferma
Chiarezza
Abbastanza chiara
Idoneità per principianti
35/100

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.