its a repo where anyone can open an issue about anything graph related and get an actual answer. subgraph devs, indexers, delegators, curators, doesnt matter which. you get a root cause if we can find one, or a workaround, or the name of whoever actually owns the problem, and we go and chase them.
the reason i built it is pretty simple. the answers to most of these questions already exist. theyre just in discord scrollback, and three weeks later theyre gone and the next person asks the same thing from scratch. ive watched the same handful of problems get solved four or five separate times by four or five different people, none of whom knew the others had done it.
so its seeded with 12 writeups, all from real questions asked in discord over the last few months. the bad indexers error map decoded properly out of the gateway source, so you can tell which indexer is actually broken and which entries are your own query being refused. the skip 20000 thing and why cursor pagination isnt just a workaround. the ipfs field that goes null at index time and never fills in because graph-node doesnt retry. one indexer quietly serving while six others return rows, attesting to it perfectly happily, which is the nastiest failure mode in the whole network imo. subgraph data that stops on a date because the contract went quiet and not because anything is stuck. the polygon bor archive snapshot situation with the actual blocks to eth_call to check whether the one youre downloading has holes in it.
three are still open including the base-reth storage v2 thing thats hitting operators right now, because i didnt want to only publish the ones with tidy endings.
what we promise is a human reply inside 48 hours and a public disposition on every issue, meaning nothing gets closed quietly. if its studio or the upgrade indexer or anything else running on edge and node infra then nobody outside e&n can fix it and we say so on day one, so at least youre not also waiting on us. handed off isnt the same as closed, we leave those open until theres an outcome.
worth being clear that this isnt meant to replace or compete with anything. e&n support handles a volume i wouldnt want to, and half of what comes in genuinely can only be fixed by them. this is the other half, the part where the answer needs somebody to sit down with a deployment id for twenty minutes, plus a place for those answers to survive afterwards. complement, not replacement.
credited everyone whose diagnosis i transcribed, by name, and most of that work isnt mine. marc-andré at ellipfra, paulieb, suntzu, maks, jb273, matthew at pinax, jerry and yash on the e&n side. if i got your name or your explanation wrong please tell me and ill fix it.
two asks. if you hit something and cant work out whose problem it is, thats exactly the thing to file. and if youre an indexer and an issue names your address, youre the fastest path to an answer and id much rather have you in the thread than my guess about your setup.
corrections wanted, i worked some of this out by reading gateway source rather than by knowing it