RPC: getBlockTime returns null for slot that hasn't been purged

[jstarry]

Problem

Developers expect that getBlockTime will return a non-null response when the requested block is still available via getConfimedBlock. The reason it can return null is because we only keep the last 5 epochs of stake info and stake info is needed for calculating stake-weighted timestamp for a block.

Proposed Solution

Only prune epochs from epoch_stakes when the last slot in an epoch is purged from the blockstore.

[mvines]

This problem is quite visible on mainnet-beta too:

$ curl -X POST -H "Content-Type: application/json" -d '{"jsonrpc":"2.0","id":1, "method":"minimumLedgerSlot"}' http://api.mainnet-beta.solana.com
{"jsonrpc":"2.0","result":5481065,"id":1}
$ solana block-time 5481065
Error: Block Not Found: slot=5481065

[mvines]

Alternatively/additionally consider adding the block-time estimate to the results from:

• https://docs.solana.com/apps/jsonrpc-api#getconfirmedtransaction
• https://docs.solana.com/apps/jsonrpc-api#getconfirmedblock
  which would probably be more compatible with a SQL-connector for longer transaction history, when we build that

[mvines]

At the same time, consider adding a block-time column in blockstore to reduce the compute performed by the getBlockTime RPC endpoint

[mvines]

Once #10093 is fixed, there only needs to be a single estimated block production time as well

[mvines]

Once blockTime is fully added to getConfirmedBlock, the getBlockTime endpoint could be deprecated and de-documented.

[mvines]

This needs #10630 to land (cc: #10984 (review))

[CriesofCarrots]

Unless we add block_time to the ConfirmedTransaction struct, I'm thinking we'll want to keep the getBlockTime endpoint, since the explorer uses it on all /tx pages. Seems like unnecessary overhead to pull the whole block for this purpose. Does this sound right, @jstarry ?

[jstarry]

Yeah the current form of the confirmed block endpoint is pretty expensive. I wouldn't mind deprecating getBlockTime if we made sure to have a lighter weight confirmed block endpoint. It could exclude tx details and paginate transaction signatures via options

[zfedoran]

Once blockTime is fully added to getConfirmedBlock, the getBlockTime endpoint could be deprecated and de-documented.

@mvines I apologize, I might not be up to date with all of the other issues you guys are solving for and this is an older issue so the conversation here might be outdated.

However, my team has a need for getting just the blockTime without any other details. We could migrate to getBlock, but it seems wasteful if all we're after is the time. Is there a recommendation to migrate to getBlock or are we okay to keep using the getBlockTime endpoint?

[mvines]

There are no current plans to deprecate getBlockTime so seems like you'd be just fine to continue using it

[github-actions]

This issue has been automatically locked since there has not been any activity in past 7 days after it was closed. Please open a new issue for related bugs.