Is ConnectParams still considered an experimental type?

[ndsistsys]

According to https://github.com/grpc/grpc-go/blame/775150f68c1759ad4b1230d09fe41efb6f3c76ad/backoff.go#L53 type ConnectParams should be considered experimental. However, the comment was added 5 years ago, can we consider this type as stable?

[eshitachandwani]

Hey, @ndsistsys , can you tell us why do you want to use it ? What is your use case ? Since we have it for a long time, there is a less chance to change it but we would still like to reserve the right to change it if necessary.

[ndsistsys]

Hey @eshitachandwani I'm interested in ConnectParams as I'd like to use some kind of backoff approach when trying to create a connection towards a gRPC server, so that a error in the first connection attempt does not make the whole process fail.
As a common strategy to avoid flooding the server with requests, I'd like to use an exponential backoff, setting up some parameters according to my use case.
I'll use a backoff strategy anyway, my doubt was using the capabilities provided by grpc-go or by another package.

Please notice that the internal backoff is mentioned at least in the following files:

grpc-go/clientconn.go

Line 609 in a51009d

// resolution, TCP connection establishment (with retries and backoff) and TLS

grpc-go/dialoptions.go

Line 284 in a51009d

// The backoff configuration specified as part of the ConnectParams overrides

Given that the definition of the struct has not change for 5 years, if you don't have anything in your backlog IMHO it would be better to delete the comment about the experimental status, as it can be misleading. At first, I understood some changes were foreseen in a near future.

[dfawley]

gRPC does exponential backoff internally automatically. Is there some reason why the default settings aren't working for you?

[ndsistsys]

I suppose the default values are

grpc-go/backoff/backoff.go

Line 47 in 6819ed7

var DefaultConfig = Config{

I have use some cases in which I need to change the MaxDelay default value.
Anyway, I'll use ConnectParams in my code base; I would like only to state again that at first, reading the comment about the struct being experimental, I understood it was not stable.
Thank you for your replies!

[dfawley]

I have use some cases in which I need to change the MaxDelay default value.

Can you explain why?

I would like only to state again that at first, reading the comment about the struct being experimental, I understood it was not stable.

Correct, it is not stable. At least in part because we haven't really found people needing to use it. I'm fairly sure we would have made it internal-only if we had known better (or maybe internal didn't exist at the time or maybe we didn't know about it?).

[github-actions]

This issue is labeled as requiring an update from the reporter, and no update has been received after 6 days. If no update is provided in the next 7 days, this issue will be automatically closed.

[ndsistsys]

@dfawley

Can you explain why?

In order to setup an optimized configuration, shouldn't we tune the backoff parameters according to a connection deadline?