Repository navigation
five RPC pages were measuring endpoints that had stopped answering (cherry-pick of #2935) - #2937
Closed
Flotapponnier wants to merge 1 commit into
Closed
Flotapponnier wants to merge 1 commit into
Flotapponnier wants to merge 1 commit into
Conversation
A sweep of the 64 thin and unread RPC pages probed all 159 configured
endpoints with the call each chain actually speaks, and 14 pages had at
least one dead endpoint. The dominant cause is not random rot: PublicNode
stopped serving five of our chains. The identical request still succeeds
on ethereum-rpc.publicnode.com and 404s on moonbeam from the same edge
with DNS resolving, so this is a withdrawal, not an outage.
Repaired here, each replacement probed with eth_getBlockByNumber or the
chain's own method, kept only after returning the same head as another
source on 10 of 10 tries:
moonbeam publicnode and tenderly 404, and the Foundation endpoint
refuses the TLS handshake from two independent stacks
(LibreSSL 3.3.6 and OpenSSL 3.5.5), so it cannot be
measured from our probes whatever the cause. Left drpc
alone, added onfinality and unitedbloc.
moonriver same PublicNode withdrawal, same Foundation refusal. Added
drpc and unitedbloc, both of which the spec had excluded in
August for a dead route and dead DNS and which answer again.
astar 1RPC answers about 15% of the time, the same ok_rate that
already got it dropped from Oasis Sapphire. Added the Astar
Foundation endpoint and drpc.
cosmos-hub Lava returns 410 Gone, consistent with its removal in #2926.
Replaced with ecostake.
gravity Ankr requires a key on this chain now and answers 403, so it
can never produce a sample. Replaced with thirdweb.
Two things found in passing and fixed:
Moonbeam's intro claimed four endpoints and carried a clause with no
subject. Moonriver's intro still listed its August cohort and named two
providers as excluded that work today. Both rewritten against what is
actually measured.
gravity-official was published with no endpoint field at all, on a page
whose own description promises URLs for every no-key endpoint. Added
rpc.gravity.xyz, the host the harness has been probing all along.
Gravity keeps its "publicnode" provider label even though that row is the
Foundation's own host. Renaming it would start a new series and blank the
column for a full day, and the page already displays it correctly.
Not recovered, recorded here so the next sweep does not re-probe them:
dymension, nibiru, conflux, merlin, swellchain, regen and nolus have no
third keyless endpoint that answers, so they stay under the three-provider
gate.
Collaborator
Author
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Cherry-pick of #2935, already merged to dev.
This has to reach
mainto have any effect: the Railway rpc-capabilitiesservices deploy from
main, so a dev-only merge changes which providers thepage declares without changing which endpoints are probed. Verified by
measurement, not assumption: after merging #2935 to dev and redeploying all
three regions with --from-source, Prometheus still showed only the old provider
set, and
maincarries the ETC repair from this morning (which did takeeffect) while it does not carry this one.
Without this merge the five pages declare providers nobody probes and publish
empty rows.
Gates on this branch: go build, go vet, pnpm validate (245 specs), and an
endpoint-match check confirming every URL the specs publish is the URL the
harness probes.