Page MenuHomePhabricator

[services] Tunnelbroker - Rust FCM push notifications library wrapper
ClosedPublic

Authored by max on Aug 22 2022, 4:50 AM.
Tags
None
Referenced Files
F3504517: D4894.diff
Fri, Dec 20, 9:22 AM
Unknown Object (File)
Mon, Dec 16, 10:12 AM
Unknown Object (File)
Mon, Dec 16, 10:12 AM
Unknown Object (File)
Mon, Dec 16, 10:12 AM
Unknown Object (File)
Mon, Dec 16, 10:12 AM
Unknown Object (File)
Sun, Dec 15, 10:39 PM
Unknown Object (File)
Sun, Dec 15, 9:11 AM
Unknown Object (File)
Sun, Dec 15, 9:11 AM

Details

Summary

This is a Rust fcm fcm notifications library wrapper to use the library in the C++ codebase to send push notifications to Android devices.

Linear task: ENG-1303

Test Plan
  1. Rust:

Run cargo build from services/tunnelbroker/rust-notifications directory.
Rust library will be successfully built.

  1. C++ (Docker):

Patch to D4807 (top of the stack) using arc patch D4807.
Running run run-tunnelbroker-service successfully built the Rust library and link it.

  1. C++ (Nix):

Patch to D4807 (top of the stack) using arc patch D4807.
Call cd services/tunnelbroker.
Running rm -dfr build && cmake -B build . && make -C build -j4 successfully built the Rust library and link it.

Library's FCM notifications push method can be called from the C++ as:

#include "rust_notifications/src/lib.rs.h"
#include "rust/cxx.h"
...

const unsigned long messageID = sendNotifToFCM(
    char const *fcmAPIKey,
    char const *deviceRegistrationID,
    char const *messageTitle,
    char const *messageBody
);

Diff Detail

Repository
rCOMM Comm
Branch
add-fcm-library-wrapper
Lint
No Lint Coverage
Unit
No Test Coverage

Event Timeline

max held this revision as a draft.
max edited the test plan for this revision. (Show Details)
max added reviewers: varun, tomek, karol, jon.
max edited the summary of this revision. (Show Details)
max published this revision for review.Aug 22 2022, 5:03 AM
max edited the test plan for this revision. (Show Details)
services/tunnelbroker/rust-notifications/src/fcm.rs
19

I don't like the usage of unwrap(), it essentially throws away a potential exception.

We should really be inspecting the result, logging if there was an error, and then not causing a panic if it does return an Error.

services/tunnelbroker/rust-notifications/src/lib.rs
75

Seems like we are going from Result<FcmResponse> to bool denoting if it was Ok or Error. I think we should just be returning the Result value.

Unless this is to be "more compatible" of a method to C++. In which the Result type doesn't really fit into the C++ "exception" landscape.

jon requested changes to this revision.Aug 22 2022, 9:49 AM
This revision now requires changes to proceed.Aug 22 2022, 9:49 AM

Use Anyhow and Result instead of bool.
Removing unwrap().
Using snake_case in Rust code, export to C++ as a camelCase.
Simplify the code.

max added inline comments.
services/tunnelbroker/rust-notifications/src/fcm.rs
19

I don't like the usage of unwrap(), it essentially throws away a potential exception.

We should really be inspecting the result, logging if there was an error, and then not causing a panic if it does return an Error.

This is a good point. I've switched to using Anyhow and Result.

services/tunnelbroker/rust-notifications/src/lib.rs
75

Seems like we are going from Result<FcmResponse> to bool denoting if it was Ok or Error. I think we should just be returning the Result value.

Unless this is to be "more compatible" of a method to C++. In which the Result type doesn't really fit into the C++ "exception" landscape.

Makes sens. I've changed it to return a Result with the message_id in case of success.

jon requested changes to this revision.Sep 1 2022, 9:32 AM
jon added inline comments.
services/tunnelbroker/rust-notifications/src/lib.rs
57–63 ↗(On Diff #16152)

can be written more minimally. Currently we're just unwrapping message_id with ?, and then re-wrapping it with Ok()

This revision now requires changes to proceed.Sep 1 2022, 9:32 AM
max marked 2 inline comments as done.

Removing of wraping result in Ok.

max added inline comments.
services/tunnelbroker/rust-notifications/src/lib.rs
57–63 ↗(On Diff #16152)

can be written more minimally. Currently we're just unwrapping message_id with ?, and then re-wrapping it with Ok()

I wasn't sure about code readability to make it this way, because it can be harder to understand what the function returns to an upper level, and the rust-analyzer highlights only future here.
But there is a return type right above, so I think it will be ok.
I've made this change.

Rust usage looks good to me

tomek requested changes to this revision.Sep 2 2022, 1:45 AM
tomek added inline comments.
services/tunnelbroker/rust-notifications/src/fcm.rs
10 ↗(On Diff #16212)

Why do we need to create a new client for every notification?

This revision now requires changes to proceed.Sep 2 2022, 1:45 AM
max marked 2 inline comments as done.
max added inline comments.
services/tunnelbroker/rust-notifications/src/fcm.rs
10 ↗(On Diff #16212)

Why do we need to create a new client for every notification?

The Client::new() is a reqwest::ClientBuilder. The HTTP request itself calls in Send. There is no point here to use a shared reqwest::ClientBuilder and reuse it because we will call this function from the different C++ threads and in this case, we will need to use a synchronization mechanism for the shared access and this will level out benefits and makes the code more complex on this simple call. Also, this can be a bottleneck in the case of the performance while parallel sending.

tomek requested changes to this revision.Sep 2 2022, 3:48 AM
tomek added inline comments.
services/tunnelbroker/rust-notifications/src/fcm.rs
10 ↗(On Diff #16212)

It seems like the main reason for not reusing the client is synchronization. Have you considered using a thread local approach?

And as a separate thing, if we expect that a couple of clients will be sending at the same time, have we tested if it actually works? What is the limit of number of concurrent clients sending at the same time?

This revision now requires changes to proceed.Sep 2 2022, 3:48 AM
max marked 2 inline comments as done.
max added inline comments.
services/tunnelbroker/rust-notifications/src/fcm.rs
10 ↗(On Diff #16212)

It seems like the main reason for not reusing the client is synchronization. Have you considered using a thread local approach?

The main reason here is that it will be a bottleneck in the case of parallel sending because we are using a synchronized single-client approach instead of scaling it according to the load. The benefit of reusing one client is overlapped by reducing scalability and adding complexity.

And as a separate thing, if we expect that a couple of clients will be sending at the same time, have we tested if it actually works?

This function is blocked until the result is came. That's why it will be called in a separate thread in C++.
Every call creates a separate connection to the FCM and it can be scaled and will not block each other.

What is the limit of number of concurrent clients sending at the same time?

There is no limit, we spawn a new connection to FCM on every Send and close it. We can maintain a limit at the upper level of the tunnelbroker where we will call it. Something like an atomic counter with the limit in Constants.

This revision is now accepted and ready to land.Sep 2 2022, 5:58 AM
max marked an inline comment as done.

Rebase and merge.

This revision was landed with ongoing or failed builds.Sep 6 2022, 5:10 AM
This revision was automatically updated to reflect the committed changes.