Improve substreams error handling #5160
Conversation
mangas
force-pushed
the
filipe/handle-substreams-errors
branch
3 times, most recently
from
e12696c to
0441af0
Compare
graph/src/blockchain/block_stream.rs
Outdated
| @@ -408,6 +409,10 @@ pub trait BlockStreamMapper<C: Blockchain>: Send + Sync { | |||
| None => return Err(SubstreamsError::MissingClockError), | |||
| }; | |||
|
|
|||
| if clock.number > 11808300 { | |||
| panic!("end block") | |||
There was a problem hiding this comment.
great catch, this was leftover from rebase I think
| ); | ||
| return; | ||
| } | ||
| _ => unreachable!(), |
There was a problem hiding this comment.
Why is BlockStreamError::Unknown not handled? would only Fatal errors be returned at this point?
There was a problem hiding this comment.
at the moment, only deterministic errors get returned, maybe I'll add some more information to this, thanks for pointing it out
There was a problem hiding this comment.
I had left a comment on process_blocks but I will add something here as well
| @@ -748,10 +748,7 @@ fn insert_subgraph_error(conn: &PgConnection, error: &SubgraphError) -> anyhow:: | |||
| } = error; | |||
|
|
|||
| let block_num = match &block_ptr { | |||
| None => { | |||
| assert_eq!(*deterministic, false); | |||
There was a problem hiding this comment.
Why was this assertion removed?
There was a problem hiding this comment.
This was discussed with the team, as far as I can understand this was a sanity check that determisnistic errors could only happen with a block number being set, this is not longer the case with substreams, since the deterministic error can even happen at boot time, for example, with incorrect number of arguments.
mangas
force-pushed
the
filipe/handle-substreams-errors
branch
from
0441af0 to
56fed4b
Compare
Code::InvalidArgumentis returned as the grpc status codeOption<String>withFirehoseCursor.