Service unavailable when fetching validator duties

Description

When fetching the validator duties for a given epoch (a recent one close to the head) I get that the service is not available. Note that the endpoint I’m using is eth/v1/validator/duties/proposer/xxx and that the node is working properly (i.e. curl http://localhost:5051/eth/v1/beacon/genesis works)

{"code":503,"message":"Service unavailable"}

Note also that I started synching the node with an Infura provided initial-state several hours before, so perhaps the node hasn’t downloaded the whole beacon chain. However, I’m fetching the duties of very recent epochs so it shouldn’t be related.

Steps to Reproduce

curl localhost:5051/eth/v1/validator/duties/proposer/143159

Expected behavior: Get the duties of this given epoch

Actual behavior: Return a “Service unavailable” error

Frequency: Always

Versions (Add all that apply)

• Software version: consensys/teku:22.8.1
• Java version: NA
• OS Name & Version: NA
• Docker Version: NA
• Cloud VM, type, size: NA

[ajsutton]

Ah, calculating proposer duties requires access to a BeaconState from within that epoch. By default Teku using prune storage mode which doesn’t store any states prior to the finalized checkpoint so when you’re requesting proposer duties for a historic state the service is unavailable because that state has been pruned and the information is no longer available. We should provide a more useful reason here though.

So you’d need to set --data-archive-mode=archive to get Teku to retain older states but it still wouldn’t have information from before you first enable that or before the initial state you sync from. To get the entire history you’d have to sync from genesis.

Alternatively you could try out the experimental support for reconstructing historic states which lets you using initial-state to get in sync quickly but then backfills the historic beacon states. It will still take quite a while to get through them all but would be faster than syncing from genesis as we don’t need to revalidate signatures. You can enable it with --Xreconstruct-historic-states.

[alrevuelta]

Thanks! I see, great explanation!

Closing, since it’s expected behavior. I solved my problem by just removing the initial-state flag and syncing the node from genesis since the suggested alternatives, archive doesn’t fully solve my issue and I don’t want to rely on experimental stuff.