graphprotocol
Feature request - signal planned maintenance
Seeing the reliability depends on % of queries sent by the gateway that don't fail, i think we need to have a way to signal planning in maintenance, which on infrastructure his elaborate and complex could easily be a few hours. Being able to do maintenance is important for the security of the servers and reliability of the indexing service.
Potentially this could also be used reactively, when you have observed a significant problem and realise you need time to remediate, to avoid further dropping in reliability score.
We probably want to combine it with a "ready for business again" to be send out when complete.
The easiest way would be to add an up (or similar) field to the /status API served by the indexer service. Gateways can then poll that periodically to decide which indexers to include in queries. This is already done in a very similar way with the /version API — indexers whose version is too old will be ignored (but rechecked periodically).
In combination with letting the cli set this to down, i reckon ? Would it make sense to make this more than binary, that you can set a UTC timestamp when planned maintenance would start ?
Depends a bit on the frequency, could be sufficient to poll (versus active signaling).