What happens when a HyperSync endpoint goes down temporarily?

[PaulRBerg]

The docs recommend using fallback RPCs in case the main HyperSync endpoint goes down.

It's not clear, however, what happens if fallback RPCs are not used. Will the indexer stop, indefinitely, and will it have to be manually restarted?

I remember we had a chat about this issue at one point @DZakh but I don't remember the details. Could you refresh my mind, please?

[DZakh]

The chain with a failing HyperSync endpoint will stop progressing until it starts responding again. We increase the polling interval quite a lot, but continue checking for it to come back. Then there are 4 scenarios:

• ordered multichain - the indexing progress will stop
• unordered multichain + historical sync + rollback on reorg - other chains will reach reorg threshold (around 200 blocks from the head) and continue progressing at this range from the head
• unordered multichain + historical sync + no rollback on reorg - other chains will continue progressing perfectly fine
• unordered multichian + realtime indexing - other chains will continue progressing perfectly fine

[PaulRBerg]

And just to confirm - if and when HyperSync gets back online, indexing will restart, right? On all chains, regardless of the ordered mode?

[DZakh]

https://docs.envio.dev/docs/HyperIndex/multichain-indexing#multichain-event-ordering

[DZakh]

Realtime mode starts automatically after all chains reach the reorg threshold. At this time we also start storing entity history and block hashes to handle reorgs and rollbacks.

[DZakh]

And just to confirm - if and when HyperSync gets back online, indexing will restart, right? On all chains, regardless of the ordered mode?

Yes, it'll continue indexing the chain.

Ordering only affects how different chains depend on each other.

[DZakh]

There are currently ordered and unordered modes, but non of them meanse chains are entirely independent. For this, we have an isolated mode in our pipeline.