Merge pull request #5141 from Dokploy/fix/monitoring-zombie-container

fix(monitoring): remove legacy container and default empty cronJob
This commit is contained in:
Mauricio Siu
2026-08-27 00:03:36 -06:00
committed by GitHub
3 changed files with 257 additions and 0 deletions
@@ -0,0 +1,236 @@
import { execSync } from "node:child_process";
import { docker } from "@dokploy/server/constants";
import { setupMonitoring } from "@dokploy/server/setup/monitoring-setup";
import { afterAll, beforeEach, describe, expect, it, vi } from "vitest";
const REAL_TEST_TIMEOUT = 120000;
const SERVICE_NAME = "dokploy-monitoring";
const TEST_IMAGE = "busybox:latest";
// Mock ONLY db-backed lookups and remote I/O. Dockerode stays real and talks to
// the local daemon, so the legacy-container cleanup is exercised for real.
vi.mock("@dokploy/server/services/server", () => ({
findServerById: vi.fn().mockResolvedValue({
serverId: "test-server",
serverType: "deploy",
sshKeyId: null, // -> getRemoteDocker returns the local docker instance
metricsConfig: {
server: {
type: "Remote",
port: 4500,
token: "test-token",
urlCallback: "http://localhost/callback",
cronJob: "0 0 * * *",
retentionDays: 2,
refreshRate: 60,
thresholds: { cpu: 0, memory: 0 },
},
containers: { refreshRate: 60, services: { include: [], exclude: [] } },
},
}),
}));
vi.mock("@dokploy/server/services/settings", () => ({
getDokployImageTag: vi.fn(() => "latest"),
}));
vi.mock("@dokploy/server/utils/docker/utils", () => ({
pullImage: vi.fn().mockResolvedValue(undefined),
pullRemoteImage: vi.fn().mockResolvedValue(undefined),
}));
vi.mock("@dokploy/server/utils/process/execAsync", () => ({
execAsync: vi.fn().mockResolvedValue({ stdout: "", stderr: "" }),
execAsyncRemote: vi.fn().mockResolvedValue({ stdout: "", stderr: "" }),
}));
const containerExists = async (name: string) => {
try {
await docker.getContainer(name).inspect();
return true;
} catch (error: any) {
if (error.statusCode === 404) return false;
throw error;
}
};
const serviceExists = async (name: string) => {
try {
await docker.getService(name).inspect();
return true;
} catch (error: any) {
if (error.statusCode === 404) return false;
throw error;
}
};
const swarmTaskNames = async () => {
const list = await docker.listContainers({ all: true });
return list
.flatMap((c) => c.Names.map((n) => n.replace(/^\//, "")))
.filter((n) => n.startsWith(`${SERVICE_NAME}.`));
};
const cleanup = async () => {
try {
await docker.getService(SERVICE_NAME).remove();
} catch {}
try {
await docker.getContainer(SERVICE_NAME).remove({ force: true });
} catch {}
for (const name of await swarmTaskNames()) {
try {
await docker.getContainer(name).remove({ force: true });
} catch {}
}
};
// Recreates the pre-v0.30.0 standalone agent stuck in a crash loop.
const createLegacyZombie = async () => {
const container = await docker.createContainer({
name: SERVICE_NAME,
Image: TEST_IMAGE,
Cmd: [
"sh",
"-c",
"echo 'Error starting metrics cleanup system: empty spec string'; exit 1",
],
HostConfig: { RestartPolicy: { Name: "always" }, NetworkMode: "host" },
});
await container.start();
};
const pullTestImage = async () => {
try {
await docker.getImage(TEST_IMAGE).inspect();
return;
} catch {}
await new Promise<void>((resolve, reject) =>
docker.pull(TEST_IMAGE, (err: any, stream: any) =>
err
? reject(err)
: docker.modem.followProgress(stream, (e: any) =>
e ? reject(e) : resolve(),
),
),
);
};
// The code under test hardcodes the production name, so this suite cannot
// namespace its fixtures. Skip rather than wipe a real agent on a Dokploy host.
const hasRealMonitoring = () => {
const query = (cmd: string) => {
try {
return execSync(cmd, { stdio: ["ignore", "pipe", "ignore"] })
.toString()
.trim();
} catch {
return "";
}
};
return (
query(
`docker service ls --filter name=${SERVICE_NAME} --format '{{.Name}}'`,
) !== "" ||
query(
`docker ps -a --filter name=^${SERVICE_NAME}$ --format '{{.Names}}'`,
) !== ""
);
};
describe.skipIf(hasRealMonitoring())(
"setupMonitoring - legacy container cleanup (real docker)",
() => {
beforeEach(async () => {
await pullTestImage();
await cleanup();
}, REAL_TEST_TIMEOUT);
afterAll(async () => {
await cleanup();
}, REAL_TEST_TIMEOUT);
it(
"removes the legacy standalone container left behind by the swarm migration",
async () => {
await createLegacyZombie();
expect(await containerExists(SERVICE_NAME)).toBe(true);
await setupMonitoring("test-server");
expect(await containerExists(SERVICE_NAME)).toBe(false);
expect(await serviceExists(SERVICE_NAME)).toBe(true);
},
REAL_TEST_TIMEOUT,
);
it(
"removes the legacy container without touching swarm tasks, which are named dokploy-monitoring.<slot>.<id>",
async () => {
const taskName = `${SERVICE_NAME}.1.qh3ldvg2h9x0test`;
const task = await docker.createContainer({
name: taskName,
Image: TEST_IMAGE,
Cmd: ["sleep", "3600"],
});
await task.start();
await createLegacyZombie();
await setupMonitoring("test-server");
expect(await containerExists(SERVICE_NAME)).toBe(false);
expect(await containerExists(taskName)).toBe(true);
const inspect = await docker.getContainer(taskName).inspect();
expect(inspect.State.Running).toBe(true);
},
REAL_TEST_TIMEOUT,
);
it(
"is idempotent when no legacy container exists",
async () => {
expect(await containerExists(SERVICE_NAME)).toBe(false);
await expect(setupMonitoring("test-server")).resolves.not.toThrow();
await expect(setupMonitoring("test-server")).resolves.not.toThrow();
expect(await serviceExists(SERVICE_NAME)).toBe(true);
},
REAL_TEST_TIMEOUT,
);
it(
"deploys the service even when removing the legacy container fails",
async () => {
const failingDocker = {
getContainer: () => ({
remove: async () => {
const error: any = new Error("device or resource busy");
error.statusCode = 500;
throw error;
},
}),
getService: docker.getService.bind(docker),
createService: docker.createService.bind(docker),
};
const remoteDocker = await import(
"@dokploy/server/utils/servers/remote-docker"
);
const spy = vi
.spyOn(remoteDocker, "getRemoteDocker")
.mockResolvedValue(failingDocker as any);
try {
await expect(setupMonitoring("test-server")).resolves.not.toThrow();
expect(spy).toHaveBeenCalled(); // guards against the spy silently not intercepting
expect(await serviceExists(SERVICE_NAME)).toBe(true);
} finally {
spy.mockRestore();
}
},
REAL_TEST_TIMEOUT,
);
},
);
@@ -21,11 +21,31 @@ const getMonitoringImage = () => {
return imageName;
};
// Swarm tasks are dokploy-monitoring.<slot>.<id>, so this only matches the
// pre-v0.30.0 standalone container. A cleanup failure must not block the deploy.
const removeLegacyContainer = async (
docker: Awaited<ReturnType<typeof getRemoteDocker>>,
serviceName: string,
) => {
try {
await docker.getContainer(serviceName).remove({ force: true });
console.log("Removed legacy monitoring container ✅");
} catch (error: any) {
if (error?.statusCode !== 404) {
console.warn(
`Could not remove legacy monitoring container: ${error?.message ?? error}`,
);
}
}
};
const deployMonitoringService = async (
docker: Awaited<ReturnType<typeof getRemoteDocker>>,
serviceName: string,
settings: CreateServiceOptions,
) => {
await removeLegacyContainer(docker, serviceName);
try {
const service = docker.getService(serviceName);
const inspect = await service.inspect();
@@ -86,6 +86,7 @@ export const serverSetup = async (
...server.metricsConfig.server,
token: token,
urlCallback: urlCallback,
cronJob: server.metricsConfig.server.cronJob || "0 0 * * *",
},
containers: server.metricsConfig.containers,
},