Do not log same errors in policer twice

[fyrchik]

Do not log same errors in policer twice. Seems useless, if we cannot select any nodes from the container, it won't change during this epoch.

окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.922Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}
окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.922Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}
окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.923Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}
окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.923Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}
окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.923Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}
окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.923Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}
окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.923Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}
окт 26 10:32:18 buky neofs-node[24920]: 2022-10-26T10:32:18.923Z        error        policer/check.go:98        could not build placement vector for object        {"component": "Object Policer", "cid": "ChimApoDVi9nYhiQohTKrVhYMvNsVuTiXMcJCr4RNyt", "error": "could not get container nodes: not enough nodes to SELECT from: ''"}

[roman-khimov]

In fact this suggests not repeating attempts as well.

[carpawell]

I think with some paraphrasing it duplicates https://github.com/nspcc-dev/neofs-node/issues/2692.

[roman-khimov]

I thought this one is a bit wider in scope, but yeah, looks like they're the same. Then the older one wins.

[cthulhu-rider]

these are different topics to me:

• this issue is about "if container nodes could not be built, skip the container until network map change"
• #2692 is about calculation caching

it'd be confusing to see how #2722 changes relate to this issue's text

[carpawell]

According to @roman-khimov's

In fact this suggests not repeating attempts as well.

i found this issue was about caching.

[cthulhu-rider]

i found this issue is about caching

caching diff stuff on different app components. I dont mind, cross-ref b/w issues exists anyway

[cthulhu-rider]

there are following node components processing storage policies:

• Object service (ACL and placement/search)
• Policer
• Used space metrics service
• Control service
• Tree service
• Reputation service (?)

all components currently process storage policies independently. This makes it possible to optimize them step by step

#2722 touches Object service as the most sensitive one in terms of performance. Other services could be done after it using a proven basis

any other thoughts?

[carpawell]

all components currently process storage policies independently

Can we unify it?