Page MenuHomePhabricator

[services] Backup - Add client reactor base classes - bidi reactor
ClosedPublic

Authored by karol on Mar 21 2022, 2:38 AM.
Tags
None
Referenced Files
Unknown Object (File)
Sat, Nov 23, 3:02 PM
Unknown Object (File)
Sat, Nov 23, 2:51 PM
Unknown Object (File)
Sat, Nov 23, 2:00 PM
Unknown Object (File)
Sat, Nov 23, 1:53 PM
Unknown Object (File)
Fri, Nov 22, 1:59 PM
Unknown Object (File)
Wed, Nov 20, 4:31 PM
Unknown Object (File)
Tue, Nov 19, 2:58 AM
Unknown Object (File)
Oct 27 2024, 4:14 AM

Details

Summary

Depends on D3513

Adding client base classes for reactors - bidi reactor

Test Plan
cd services
yarn run-backup-service

Diff Detail

Repository
rCOMM Comm
Lint
No Lint Coverage
Unit
No Test Coverage

Event Timeline

Harbormaster returned this revision to the author for changes because remote builds failed.Mar 21 2022, 2:51 AM
Harbormaster failed remote builds in B7486: Diff 10541!
karol retitled this revision from [services] Backup - Add client reactor base classes to [draft] [services] Backup - Add client reactor base classes.Mar 21 2022, 3:09 AM
karol edited the summary of this revision. (Show Details)

empty - retrigger CI

karol retitled this revision from [draft] [services] Backup - Add client reactor base classes to [draft] [services] Backup - Add client reactor base classes - bidi reactor.
karol edited the summary of this revision. (Show Details)
tomek requested changes to this revision.Mar 24 2022, 5:46 AM
tomek added inline comments.
services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
11 ↗(On Diff #10653)

Do we access this from multiple threads? Does it need to be atomic?

It might be a good idea to investigate reactors in context of threading. It may be the case that we don't need to make this atomic and library takes care of synchronizing the memory, but it is also possible that we need to do that and currently have a lot of subtle bugs.

45–48 ↗(On Diff #10653)

Shouldn't we initialize as the first operation? It sounds like a good idea to do that before calling StartWrite

81–83 ↗(On Diff #10653)

Do we need this if? terminate checks if this->done

This revision now requires changes to proceed.Mar 24 2022, 5:46 AM
services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
11 ↗(On Diff #10653)

I agree – we should have a really strong idea of how threading works with reactors

services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
11 ↗(On Diff #10653)

I agree as well. I wanted to get the grpc's authors perspective on this first if possible: https://linear.app/comm/issue/ENG-914/take-care-of-the-threading-problem-in-the-grpc-async-api

If I get no response, I'm going to investigate this on my own.

45–48 ↗(On Diff #10653)
81–83 ↗(On Diff #10653)

You're right it is redundant.

remove redundant condition, fix done callback

tomek requested changes to this revision.Mar 25 2022, 8:10 AM
tomek added inline comments.
services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
45–48 ↗(On Diff #10653)

Thanks for the clarification. It looks like term initialized confused me as initialization usually happens before any operation - maybe we can find a less confusing name? Alternatively, in the examples you linked, this call happens in the constructor. Is it possible to use the same approach here?

This revision now requires changes to proceed.Mar 25 2022, 8:10 AM
karol added inline comments.
services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
45–48 ↗(On Diff #10653)

This cannot be done, because prepareRequest is being called here and it is a pure virtual function. What happens when we create an object of a derived class is, first the constructor of the base class is being called. And in the base class, we have only the pure virtual version of this method, therefore we have to call nextWrite (and effectively prepareRequest) explicitly on an object of the derived class.

tomek added inline comments.
services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
45–48 ↗(On Diff #10653)

Ok, that makes sense

This revision is now accepted and ready to land.Mar 28 2022, 5:18 AM
karol retitled this revision from [draft] [services] Backup - Add client reactor base classes - bidi reactor to [services] Backup - Add client reactor base classes - bidi reactor.

Removing the "draft" label.

tomek added inline comments.
services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
58–59 ↗(On Diff #10747)

Is that the case that when status is non null, done is always true? If that's the case, maybe we don't need done and in isDone we can return this->status != null

72 ↗(On Diff #10747)

This can be probably achieved by calling get method on a shared pointer - for me it's less confusing, but up to you: if you prefer the current approach we can keep it

This revision is now accepted and ready to land.Mar 29 2022, 8:39 AM
services/backup/docker-server/contents/server/src/Reactors/client/base-reactors/ClientBidiReactorBase.h
58–59 ↗(On Diff #10747)

The problem is, this->status cannot be nullptr because it is not a pointer, just a normal object of a class. So I guess it has some initial values(like error code 0). We could keep status as a smart pointer, but I'd keep it consistent across all reactor base classes so if you feel we need this, I can create a task and do it later.

72 ↗(On Diff #10747)

I don't mind, no difference for me(even in readability honestly).