Add error message for eth_getTransactionReceipt when ancient blocks haven't been downloaded yet (warp-sync)

[jimpo]

I'm running:

• Which Parity version?: 2.0.0
• Which operating system?: Linux
• How installed?: from source
• Are you fully synchronized?: yes
• Which network are you connected to?: ethereum
• Did you try to restart the node?: yes

A warp-synced full node returns empty responses to eth_getTransactionReceipt RPC requests for transactions before the snapshot point.

# Request transaction from recent block
$ curl -H 'Content-Type: application/json' --data '{"jsonrpc":"2.0","method":"eth_getTransactionReceipt","params":["0xf8e026115bafda9efaee095d1f3342333826f53e332d4ffd7c9fb52bc499f9d5"],"id":0}' localhost:8545
{"jsonrpc":"2.0","result":{"blockHash":"0x7d24546c087fe34a595a250ba3aa3b4fa8c6cbfa20ec3963e122b1234170742b",...

# Request transaction from old block
$ curl -H 'Content-Type: application/json' --data '{"jsonrpc":"2.0","method":"eth_getTransactionReceipt","params":["0x511b14d534466fdc5163d2cd7e37910a570c8e012e9970c7206272b4f8b122c5"],"id":0}' localhost:8545
{"jsonrpc":"2.0","result":null,"id":0}

I assume Parity does not back fill transaction receipts from ancient blocks in warp mode, but the current behavior seems broken. The RPC should either return an error if the receipt is missing or fetch it on demand.

[Tbaut]

Was your node fully synced (with blocks older than the snapshot)?
Warp-sync is a way to get in sync quickly, but the node will download and verify all the ancient blocks once the sync is performed.
The logs looks like this (see Verifying blocks after a warp sync) https://wiki.parity.io/FAQ#what-does-paritys-command-line-output-mean

[jimpo]

Hmm, my ancient block download seems to be getting stuck at #54. I'll have to look into that separately.

I didn't know about the ancient block backfill, that makes sense. Thanks! But the RPC should still return an error saying "The receipt may or may not exist, but we won't know until we're done syncing" rather than null for the receipt.

[Tbaut]

A better error message would be nice yes. I edited your title.

[olekon]

Is there a way to download only event logs after warp sync?
Or is it possible to warpsync only to specific block and then switch to full sync?

[Tbaut]

No it's not possible. a full sync needs something like a couple days depending on your hardware.

[olekon]

Is it possible to track the progress of this backfill? The block numbers in logs are unordered
47690, then 53915, then 34549

2018-07-24 21:33:01      #47690   63/100 peers    459 MiB chain  106 MiB db  0 b
ytes queue    5 MiB sync  RPC:  1 conn,  2 req/s, 197 µs
...
2018-07-24 21:34:01  Imported #6023111 0x93ec.7a4f (98 txs, 7.99 Mgas, 569 ms, 2
9.18 KiB)
2018-07-24 21:34:02      #53915   63/100 peers    469 MiB chain  105 MiB db  0 b
ytes queue    5 MiB sync  RPC:  1 conn,  2 req/s, 197 µs
2018-07-24 21:34:11  Syncing #6023111 0x93ec.7a4f     0 blk/s    0 tx/s   0 Mgas
/s      0+    0 Qed  #6023111   62/100 peers    475 MiB chain  105 MiB db  0 byt
es queue    5 MiB sync  RPC:  1 conn,  2 req/s, 265 µs
...
2018-07-24 21:35:49      #34549   66/100 peers    600 MiB chain  106 MiB db  0 b
ytes queue    6 MiB sync  RPC:  1 conn,  2 req/s, 172 µs

[Tbaut]

@olekon what you experience (unordered download of older blocks) is not related to the OP and looks like a bug. Could you please open a dedicated issue?

[olekon]

@Tbaut I've opened #9212 earlier but you asked me to comment here for some reason

Anyway. I've got that only full sync from scrath will help me to get event logs from past blocks

[Tbaut]

#9212 is about logs not available for older blocks, that's fine. Now we shouldn't discuss anything else such as "the block numbers in logs are unordered". This is a different issue that I ask you to open :)

[andresilva]

Related #7411.

[tomusdrw]

Might be quite difficult to implement, would require the warp snapshot to actually contain all transaction hashes that are expected to be in the ancient blocks.

I think much better approach in general would be to fallback to light-like mode, so if we don't find a transaction hash and we know that we warped we query peers for that hash.