Push to backend needs a sanity check before performing

[sudheesh001]

• Issue type: Feature request / Enhancement

So this proposal is broken into separate parts which need to be implemented 1 after the other for this to be completed.

1. Improve the response from the hello.json to take a value from config called accept-push-messages: true. If this flag is set to be true, publish the space available in the backend in the response to hello.json
2. Caretaker threads currently fail to push via the push.json api to the backend servers (main peer) either when the disk runs out space or when the servers are down. Performing a hello.json check and finding out disk space before pushing the messages seems to be a good option. This will also ensure that if the server is down at any point, only a failed socket error is thrown, instead of failed message push stack traces.

Some thought experiments:

Can we use a new flag in config like player: junior|senior above to be published to the main backend peer to register the IP address of the peer in peers.json. A local instance running on someones PC could fetch these tables of registered network nodes periodically and push to them in round robin fashion or if a server denies a push for some reason?

[mariobehling]

Yes, we had a discussions about a similar implementation earlier. Sounds like it can work well for me.

[singhpratyush]

This is dependent on #1460.