Skip to content

Conversation

@lutter
Copy link
Collaborator

@lutter lutter commented Apr 24, 2025

Unassigning subgraphs has a number of unwanted side-effects, e.g., canceling copies. Pausing is safer alternative to avoid spending cycles on a subgraph we know will not make progress.

@lutter lutter requested a review from incrypto32 April 28, 2025 18:09
Unassigning subgraphs has a number of unwanted side-effects, e.g.,
canceling copies. Pausing is safer alternative to avoid spending cycles on
a subgraph we know will not make progress.
@lutter lutter force-pushed the lutter/pause-on-failure branch from 49ac66c to 12c91aa Compare May 1, 2025 18:03
@lutter lutter merged commit 12c91aa into master May 1, 2025
@lutter lutter deleted the lutter/pause-on-failure branch May 1, 2025 18:04
cjorge-graphops added a commit to graphops/graph-node that referenced this pull request Nov 14, 2025
After this was merged graphprotocol#5160 substreams fatal errors can write SubgraphError{ block_ptr: None }. This introduces a bug that shows in Status API as those records can surface as fatalError.block = null violating the schema of type Block { hash: Bytes! number: BigInt! }

A later commit graphprotocol#5971 makes it much easier to trigger this when doing a status API that returns all subgraphs (empty [] array for deployment ID) as failed substreams linger on in the indexing status set as paused instead of being unassigned.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants