Launch runbook — basebid.lol launchpad

Working copy: this repo (gitlawb-platform/openlaunch) is the single source since 2026-09-07. Deploy from app/ with the command below; the old basebid deploy tree is archived. Local-only files that never enter git: app/.env.local, app/.env.development.local, app/.env.production, contracts/.env, contracts/.launchpad-deployer.address.

0. Deploy the contracts (Base mainnet)

Deployer key lives in the macOS keychain (service basebid-launchpad-deployer); address in ../contracts/.launchpad-deployer.address. Fund it with ~0.001 ETH on Base (deploy ≈ 0.00006 ETH).

cd ../contracts
DEPLOYER_PRIVATE_KEY=$(security find-generic-password -s basebid-launchpad-deployer -w) \
  forge script script/DeployLaunchFactory.s.sol --rpc-url https://mainnet.base.org --broadcast
# → LaunchFactory + LaunchLocker addresses + the deploy block (broadcast/…/run-latest.json)
forge verify-contract <factory> src/LaunchFactory.sol:LaunchFactory --chain base \
  --constructor-args $(cast abi-encode "c(address,address,address)" 0x498581fF718922c3f8e6A244956aF099B2652b2b 0x7C5f5A4bBd8fD63184577525326123B519429bDc 0x000000000022D473030F116dDEE9F6B43aC78BA3)
forge verify-contract <locker> src/LaunchLocker.sol:LaunchLocker --chain base \
  --constructor-args $(cast abi-encode "c(address)" 0x7C5f5A4bBd8fD63184577525326123B519429bDc)

1. Wind down the old board (owner = the BaseBid deployer EOA)

The old escrow (0x7a6373BC4796971059824D5ce8DF71AdE54d0285) still holds bidders' USDC and the withdraw UI is gone. Push every refund from the owner key, then nothing else is needed:

cast send 0x7a6373BC4796971059824D5ce8DF71AdE54d0285 "endSeasonEarly()" --rpc-url https://mainnet.base.org --private-key $OWNER
cast send 0x7a6373BC4796971059824D5ce8DF71AdE54d0285 "batchRelease(uint256,uint256)" 1 100 --rpc-url https://mainnet.base.org --private-key $OWNER
# check: cast call … "totalLocked()(uint256)" → 0

(function names per ../contracts/src/BaseBid.sol; see ../contracts/docs/BASEBID.md.)

2. Fly

cd basebid
NEXT_PUBLIC_LAUNCH_FACTORY=0x… NEXT_PUBLIC_LAUNCH_LOCKER=0x… LAUNCH_DEPLOY_BLOCK=… ./scripts/fly-secrets.sh
fly deploy -c fly.toml --ha=false --remote-only \
  --build-arg NEXT_PUBLIC_LAUNCH_FACTORY=0x… --build-arg NEXT_PUBLIC_LAUNCH_LOCKER=0x…
curl -s https://basebid.lol/api/health

The release command applies db/schema.sql (adds the launch tables; leaves the old bb_* board tables in place — drop them by hand later if wanted).

3. Smoke

Run scripts/smoke.sh after EVERY deploy (all page types + live APIs must be 200). Then: Launch one token from the site with a real wallet (0% fee), buy 0.001 ETH of it, sell half, press Collect on a 1% launch. Check /api/launch/feed shows all of it.

Token logo uploads (Tigris bucket)

Logos are uploaded through POST /api/launch/image (wallet + per-IP rate limits, magic-byte sniff, sharp re-encode to a 512² WebP with metadata stripped, random key) into a PUBLIC Tigris bucket. The app only ever PUTs; nothing is deletable through the site. Create the bucket once (stages the AWS_* + BUCKET_NAME secrets on the app automatically):

fly storage create -a basebid -n openlaunch-images --public

Then fly deploy. Public URL base defaults to https://<BUCKET_NAME>.fly.storage.tigris.dev (override with IMAGE_PUBLIC_BASE). Without the bucket the route answers 503 and the form falls back to the paste-a-URL field. Local dev: IMAGE_STORE=local → .uploads/ served by /api/launch/image/<key>. The per-token share card embeds a logo ONLY when it lives on that bucket host (isOwnImageUrl).

Database backups (Fly built-in, enabled 2026-09-07)

Two layers, both automatic:

Volume snapshots — Fly snapshots the basebid-db volume daily; retention raised to 30 days (fly volumes update <vol> -a basebid-db --snapshot-retention 30). Restore = new volume from the snapshot → new Postgres machine. Day granularity.
Postgres backups to Tigris (fly pg backup enable -a basebid-db, then fly secrets deploy -a basebid-db to apply — that restarts the single Postgres machine, ~2 min of DB downtime): WAL archived every 60s, full base backup every 24h, 7-day point-in-time recovery window, bucket basebid-db-postgres (private, managed by Fly). Check: fly pg backup list -a basebid-db · fly pg backup config show -a basebid-db.

Restore rehearsal (passed 2026-09-07 01:00Z): fly pg backup restore basebid-db-restore-test -a basebid-db --restore-target-name <ID> built a fresh cluster in ~100s; row counts matched production up to the backup's end time (333 launches / 7,794 swaps / 522 metadata rows / 8 posts / 237 fee events). Query a restored cluster with fly ssh console -a <app> -C "sh -c 'PGPASSWORD=\$OPERATOR_PASSWORD psql -h localhost -p 5433 -U postgres -d basebid -c ...'" (port 5433, not the socket). Destroy the rehearsal app afterwards. For a real recovery: restore to a new app, fly postgres detach/attach it to basebid (DATABASE_URL), redeploy, run scripts/smoke.sh. Point-in-time: --restore-target-time <RFC3339> instead of a name.

5.1 KiB Raw Permalink Blame History