fix(backend): submit playground classic txs via Horizon, not Soroban RPC#39
Merged
Merged
Conversation
The playground built and confirmed its classic mess transactions through the load-balanced public Soroban RPC. Its nodes lag each other and the just-closed ledger, which surfaced as a rotating cast of failures on random steps: account-not-found on the first read, tx_bad_seq (stale sequence), tx_no_account (source not yet ingested on the node that received the submit), and 90s+ pollTransaction confirmation timeouts. All are the same root cause - reading and confirming classic state against an inconsistent RPC. The rest of the codebase already treats Horizon as the lag-free source of truth for classic accounts (account-live.ts). Horizon also submits classic txs synchronously - it holds the request until the tx is included - so there's no separate read-then-poll window for the lag to slip through. - route buildSignSubmit's account reads and submission through Horizon (lib/playground/horizon-submit.ts, HORIZON_URLS) instead of the Soroban RPC - broaden the (now belt-and-braces) retry to the transient RPC-lag codes tx_bad_seq + tx_no_account; rename BAD_SEQ_* to SUBMIT_RETRY_* - raw fetch, not the SDK Horizon Server, to avoid a second SDK copy (v16 dual-build hazard) - tests for tx_no_account retry and the renamed budget Verified on testnet: 4/4 sessions (3 Full at 13 steps + 1 Standard) completed every mess step, ~4-6s each, with zero bad_seq / no_account / timeout - up from 1/4 against the Soroban RPC under the same conditions.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
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
1 participant
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.
Problem
After the earlier retry fixes, the playground still failed on random mess steps in production. Driving the API directly surfaced a rotating cast of failures, all from the load-balanced public Soroban RPC whose nodes lag each other and the just-closed ledger:
Account not foundon the first read (ingestion lag)tx_bad_seq— stale sequence served to a consecutive same-account submittx_no_account— SETUP confirmed, but the next submit hit a node that hadn't ingested the source yetpollTransactionconfirmation timeoutsMeasured live: the Soroban RPC was running ~3 ledgers (~15s) behind Horizon, and a test submit returned ERROR. A direct API run scored 1/4 full sessions.
These are one root cause: reading and confirming classic state against an inconsistent RPC. The codebase already treats Horizon as the lag-free source of truth for classic accounts (
account-live.ts), and Horizon submits classic txs synchronously (it holds the request until inclusion), so there's no read-then-poll window for the lag to slip through.Changes
buildSignSubmitnow reads the source account and submits through Horizon (lib/playground/horizon-submit.ts,HORIZON_URLS) instead of the Soroban RPC. Does not touch the main demolish flow.tx_bad_seq+tx_no_account;BAD_SEQ_*renamed toSUBMIT_RETRY_*.fetch, not the SDK HorizonServer, to avoid a second SDK copy (v16 dual-build hazard).tx_no_accountretry and the renamed budget.Verification
bun type-check✅ ·bun lint✅ ·bun test tests/unit✅ (247 pass)bad_seq/no_account/ timeout — up from 1/4 against the Soroban RPC under the same conditions.