Give agent sandboxes time to start after a deploy (#22)

The first sandbox on a host after a deploy pulls the runner image, which
takes longer than the containers library's 8 s and 20 s defaults, so
creating a sandbox returned 500. Wait up to three minutes, and if the
container still doesn't start, mark the sandbox failed and return 503.


Claude-Session: https://claude.ai/code/session_01AizKXSsengXZRegJxRATkL

Co-authored-by: Claude <noreply@anthropic.com>

Canberk Demirci authored 13 days ago· GitHub committed 13 days agosigned1 parent d10e3dbcommit 388e5b48f001fd71fe0e030777e412f1fa783ae3Browse files

3 files changed+26−13

modified.github/workflows/live-test.yml+1−1View file

modifiedapps/diffinery/src/repo/actions.ts+18−11View file

modifiedapps/diffinery/src/runner/containers.ts+7−1View file

		@@ -283,7 +283,7 @@
283	283	run: \|
284	284	set -euo pipefail
285	285	base="$DIFFINERY/api/v3/repos/$OWNER/$REPO/diffinery/sandboxes"
286		- id=$(curl -fsS -X POST "$base" -H "Authorization: Bearer $TOKEN" -H "content-type: application/json" --data '{}' \| jq -r .id)
	286	+ id=$(curl -fsS --retry 1 --retry-delay 15 -X POST "$base" -H "Authorization: Bearer $TOKEN" -H "content-type: application/json" --data '{}' \| jq -r .id)
287	287	echo "sandbox $id"
288	288	for i in $(seq 1 40); do
289	289	out=$(curl -sS -X POST "$base/$id/exec" -H "Authorization: Bearer $TOKEN" -H "content-type: application/json" --data '{"cmd":"cat README.md && git lfs ls-files \| wc -l"}' \|\| true)

		@@ -1310,17 +1310,24 @@
1310	1310	const secret = randomToken(32);
1311	1311	this.sql.exec("INSERT INTO sandboxes (id, secret, created_by, ref, created_at, status) VALUES (?, ?, ?, ?, ?, 'running')", id, await sha256(token), JSON.stringify(caller), ref ?? m.defaultBranch, now());
1312	1312	const stub = this.env.SANDBOX.get(this.env.SANDBOX.idFromName(id));
1313		- await stub.boot({
1314		- DIFFINERY_CLONE_URL: `${this.env.PUBLIC_URL}/${m.owner}/${m.name}.git`,
1315		- DIFFINERY_REF: ref ?? m.defaultBranch,
1316		- DIFFINERY_JOB_TOKEN: token,
1317		- DIFFINERY_SANDBOX_SECRET: secret,
1318		- DIFFINERY_API_URL: this.env.PUBLIC_URL,
1319		- GIT_AUTHOR_NAME: caller.login,
1320		- GIT_AUTHOR_EMAIL: `${caller.login}@users.noreply.diffinery.com`,
1321		- GIT_COMMITTER_NAME: caller.login,
1322		- GIT_COMMITTER_EMAIL: `${caller.login}@users.noreply.diffinery.com`,
1323		- }, secret);
	1313	+ try {
	1314	+ await stub.boot({
	1315	+ DIFFINERY_CLONE_URL: `${this.env.PUBLIC_URL}/${m.owner}/${m.name}.git`,
	1316	+ DIFFINERY_REF: ref ?? m.defaultBranch,
	1317	+ DIFFINERY_JOB_TOKEN: token,
	1318	+ DIFFINERY_SANDBOX_SECRET: secret,
	1319	+ DIFFINERY_API_URL: this.env.PUBLIC_URL,
	1320	+ GIT_AUTHOR_NAME: caller.login,
	1321	+ GIT_AUTHOR_EMAIL: `${caller.login}@users.noreply.diffinery.com`,
	1322	+ GIT_COMMITTER_NAME: caller.login,
	1323	+ GIT_COMMITTER_EMAIL: `${caller.login}@users.noreply.diffinery.com`,
	1324	+ }, secret);
	1325	+ } catch (e) {
	1326	+ console.error("sandbox boot failed", e);
	1327	+ this.sql.exec("UPDATE sandboxes SET status = 'failed' WHERE id = ?", id);
	1328	+ await stub.shutdown().catch(() => {});
	1329	+ throw new HttpError(503, "the sandbox container did not start; try again");
	1330	+ }
1324	1331	return { id, ref: ref ?? m.defaultBranch };
1325	1332	}
1326	1333

		@@ -50,7 +50,13 @@
50	50
51	51	async boot(envVars: Record<string, string>, secret: string): Promise<void> {
52	52	await this.ctx.storage.put("secret", secret);
53		- await this.startAndWaitForPorts({ ports: 8080, startOptions: { envVars, entrypoint: ["node", "/opt/diffinery/agent.mjs", "--sandbox"] } });
	53	+ // The first start on a host after a deploy pulls the runner image, which takes
	54	+ // well over the library's 8 s / 20 s defaults.
	55	+ await this.startAndWaitForPorts({
	56	+ ports: 8080,
	57	+ startOptions: { envVars, entrypoint: ["node", "/opt/diffinery/agent.mjs", "--sandbox"] },
	58	+ cancellationOptions: { instanceGetTimeoutMS: 180_000, portReadyTimeoutMS: 120_000, waitInterval: 500 },
	59	+ });
54	60	}
55	61
56	62	/** Forward an API call (JSON text) to the sandbox's HTTP server with its secret attached. */

codician/diffinerypublic

Give agent sandboxes time to start after a deploy (#22)