apns - handling error from previous batch

[denisborovnev]

Sometimes 750ms is not enough to get response from apns when there is an error in batch

In this case all messages in batch marked as sent and connection is reused for new batches.

During sending next batch error is read, but there is no failed notification in current batch and reader does nothing

As a result notifications from this batch are not marked as sent or failed. Also next batches won't be sent because "error" connection is not closed and ApnsConnection instance continues to use it.

---

To fix this problem:

• socketCanWrite -> return false if there are available bytes to read in current socket, because it's error from previous batch and new notifications wroten to this connection won't be sent
• if reader can't find index of failed notification -> push notifications again to queue and close connection
• added ResponseWaitTimeout into configuration to make it's possible to increase timeout for waiting response from apns (if 750ms is not enough too often)

---

PS: there is already setting called MillisecondsToWaitBeforeMessageDeclaredSuccess, not sure if it might be used instead of new one ResponseWaitTimeout

[Redth]

@denisborovnev Is this working well for you in production?

I do think we should use the existing MillisecondsToWaitBeforeDeclaredSuccess as the value. If you can update your PR I'm happy to merge this.

Thanks :)

[denisborovnev]

Hi @Redth !
Thanks for review!

I do think we should use the existing MillisecondsToWaitBeforeDeclaredSuccess as the value.

done

Is this working well for you in production?

We haven't released yet, so can't give answer right now :) But we have already tested and haven't faced problem when server stops send any notification for some period of time.

[PashaPash]

Hi @Redth. We have that fix in production, working well, no issues yet. Could you please merge it?