Fix/Unwrap incomplete object put errors

[carpawell]

Closes #2092.

Before:

Store lock object in NeoFS: client failure: status: code = 1024 message = incomplete object PUT by placement: could not close object stream: (*putsvc.remoteTarget) could not put object to [/dns4/s02.neofs.devenv/tcp/8080]: write object via client: status: code = 1024 message = incomplete object PUT by placement: could not close object stream: could not lock object from lock objects locally: status: code = 2051
...
rpc error: remove object via client: status: code = 1024 message = incomplete object PUT by placement: could not close object stream: (*putsvc.remoteTarget) could not put object to [/dns4/s02.neofs.devenv/tcp/8080]: write object via client: status: code = 1024 message = incomplete object PUT by placement: could not close object stream: could not delete objects from tombstone locally: lock object removal
...
rpc error: remove object via client: status: code = 1024 message = incomplete object PUT by placement: could not close object stream: (*putsvc.remoteTarget) could not put object to [/dns4/s01.neofs.devenv/tcp/8080]: write object via client: status: code = 1024 message = incomplete object PUT by placement: could not close object stream: could not delete objects from tombstone locally: status: code = 2050

After:

Store lock object in NeoFS: client failure: status: code = 2051 message = locking non-regular object is forbidden
...
rpc error: remove object via client: status: code = 1024 message = lock object removal
...
rpc error: remove object via client: status: code = 2050 message = object is locked

Not sure if we want to lose error chain but that is how unwrapping (looking for API status errors) works in the current code base.

[sami-nspcc]

I have recieved a notification of a new pull request. I am starting the build of images and binaries for further testing. Build number is 3763

[codecov]

Codecov Report

Merging #2126 (49074a0) into master (42554a9) will decrease coverage by 0.25%.
The diff coverage is n/a.

❗ Current head 49074a0 differs from pull request most recent head 4c88582. Consider uploading reports for the commit 4c88582 to get more accurate results

@@            Coverage Diff             @@
##           master    #2126      +/-   ##
==========================================
- Coverage   30.78%   30.53%   -0.26%     
==========================================
  Files         382      381       -1     
  Lines       28056    28003      -53     
==========================================
- Hits         8637     8550      -87     
- Misses      18687    18721      +34     
  Partials      732      732              

Impacted Files	Coverage Δ
pkg/local_object_storage/writecache/get.go	0.00% <0.00%> (-75.87%)	⬇️
pkg/local_object_storage/writecache/init.go	45.00% <0.00%> (-48.55%)	⬇️
..._object_storage/blobstor/blobovniczatree/exists.go	77.77% <0.00%> (-6.23%)	⬇️
...object_storage/blobstor/blobovniczatree/control.go	68.49% <0.00%> (-5.48%)	⬇️
cmd/neofs-node/morph.go	0.00% <0.00%> (ø)
pkg/morph/client/notary.go	0.00% <0.00%> (ø)
pkg/local_object_storage/writecache/options.go	56.09% <0.00%> (ø)
pkg/local_object_storage/blobovnicza/exists.go
pkg/local_object_storage/writecache/storage.go	27.14% <0.00%> (+0.75%)	⬆️
pkg/local_object_storage/blobstor/exists.go	76.47% <0.00%> (+14.56%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[sami-nspcc]

I am running integration tests

[sami-nspcc]

Test run is finished. Please download the tarball from link. Untar and use allure open to watch the report

[fyrchik]

Can you give an example of what message is expected now? Is it still obvious that there was some error on others node.

[carpawell]

With that PR error messages would be just the deepest error because the current error messaging is designed to be so

I agree that keeping the chain could be a good error context and that is why im not sure about the PR. The problem that PR solves is described in the linked issue: any SDK error that has not been converted to its V2 form does not produce a text message (wrapping an SDK error by an error that does not have Unwrap method leads to calling Error method without conversion to a V2 struct => no error message). I did not like such a solution but here we are.

If our desire to save error chain in an "incomplete" context is really strong, I may suggest using init() in SDK for predefined errors that fills all error messages before their usage. Another solution could be introducing a new "IncompleteObjectPut" API error that wraps any error and has a detailed description of error numbers and reasons\

/cc @cthulhu-rider

[cthulhu-rider]

For now I don't mind

[sami-nspcc]

I have recieved a notification of a new pull request. I am starting the build of images and binaries for further testing. Build number is 3770

[sami-nspcc]

I am running integration tests

[sami-nspcc]

Test run is finished. Please download the tarball from link. Untar and use allure open to watch the report

[carpawell]

Closing this in favor of nspcc-dev/neofs-sdk-go#369.