Compare commits

...

12 Commits

Author SHA1 Message Date
autofix-ci[bot] ebeb95cefd [autofix.ci] apply automated fixes 2026-08-27 06:42:22 +00:00
Mauricio Siu 2d8dfee9f3 fix(server): keep ${VAR} self-references interpolatable in .env files
Only escape $ signs that are not part of a well-formed ${VAR} reference
when writing .env files, so Docker Compose's own variable interpolation
still works for legitimate ASSET_URL=${APP_URL}-style references, while
literal dollar signs (e.g. in passwords) stay escaped.

Fixes #5151
2026-08-27 00:41:17 -06:00
Mauricio Siu 3c6a96a30a Merge pull request #5141 from Dokploy/fix/monitoring-zombie-container
fix(monitoring): remove legacy container and default empty cronJob
2026-08-27 00:03:36 -06:00
Narciso E. Núñez Arias 546686ea35 Merge pull request #5190 from somuai/fix/allow-plus-at-in-read-valid-directory
fix(server): allow '+' and '@' in readValidDirectory path validation
2026-08-26 12:07:29 -04:00
autofix-ci[bot] 217ca07842 [autofix.ci] apply automated fixes 2026-08-26 15:05:57 +00:00
somuai bd8ba128cd fix(server): allow plus, at, and valid path characters in readValidDirectory 2026-08-26 04:54:13 +05:30
Mauricio Siu 72a554de5f chore: add script to spawn agent worktree with environment setup 2026-08-25 01:29:51 -06:00
Mauricio Siu cda047feb2 chore: add install-worktree-deps.sh script for dependency installation 2026-08-25 01:20:10 -06:00
Mauricio Siu feccb51287 Merge pull request #5183 from Dokploy/chore/worktree-automation
chore: automate git worktree setup
2026-08-25 01:03:38 -06:00
Narciso 875baa139c mend 2026-08-24 20:59:30 -04:00
Narciso e2ab2eb6bc remove unused code 2026-08-24 20:52:35 -04:00
Narciso bee6918d9a fix(monitoring): remove legacy container and default empty cronJob
The swarm migration in 3848fa9c0 dropped the container.remove({force:true})
that the standalone deploy path used to run. Swarm tasks are named
dokploy-monitoring.<slot>.<id>, so there is no name collision and the
pre-v0.30.0 container survives every redeploy. It also stays pinned to an
orphaned image ID once pullRemoteImage moves the latest tag, so neither a
pull nor a Save clears it and it restarts forever.

Cloud setup spread metricsConfig straight from the row, shipping
cronJob: "" to the agent. robfig/cron rejects an empty spec, so the Go
binary exits before Fiber binds 4500 and Docker restarts it every ~60s.

- remove the legacy container in deployMonitoringService, which covers both
  setupMonitoring and setupWebMonitoring. Cleanup is best effort: a failure
  is logged and the deploy continues, matching the pre-migration behaviour
- default cronJob when configuring monitoring for cloud
- on build servers, clean up the legacy container but deploy no service.
  They never join the swarm, yet cloud setup did create the standalone
  container there before v0.30.0, and the monitoring form has always been
  hidden for them, so those agents are all stuck with an empty cronJob
- cover the above with real-docker tests
2026-08-20 11:17:41 -04:00
13 changed files with 478 additions and 6 deletions
+5 -2
View File
@@ -1,13 +1,16 @@
{
"worktree": {
"baseRef": "fresh",
"symlinkDirectories": ["node_modules"]
"baseRef": "fresh"
},
"hooks": {
"PostToolUse": [
{
"matcher": "EnterWorktree",
"hooks": [
{
"type": "command",
"command": "\"$CLAUDE_PROJECT_DIR/scripts/install-worktree-deps.sh\""
},
{
"type": "command",
"command": "\"$CLAUDE_PROJECT_DIR/scripts/assign-worktree-port.sh\""
+17 -1
View File
@@ -4,7 +4,23 @@ description: Implement a GitHub issue with reproduction and verification
allowed-tools: Bash, Edit, Write, Read, Glob, Grep, mcp__playwright__*, mcp__dokploy__*
---
The issue number is passed as $1. An isolated instance is running at $DOKPLOY_BASE_URL.
The issue number is passed as $1.
## Instance
No instance is running yet — start your own, isolated to this worktree:
1. Check `apps/dokploy/.env` for `PORT` (assigned per-worktree already).
2. If nothing is listening on that port, start it: `pnpm dokploy:dev` in the
background, then poll `curl -s -o /dev/null -w '%{http_code}' http://localhost:$PORT`
until it answers (usually ~10-15s).
3. Use `http://localhost:$PORT` as the base URL for Playwright navigation.
Note: `mcp__dokploy__*` (this repo's `.mcp.json`) resolves its URL from
`$DOKPLOY_BASE_URL` once, at session startup — it cannot pick up a port
discovered mid-session. If those tools are unavailable or point at the wrong
instance, fall back to `curl`/`gh api` for API-level checks, or ask the user
to relaunch with `DOKPLOY_BASE_URL` exported first.
## Tools
@@ -93,4 +93,46 @@ describe("getCreateEnvFileCommand", () => {
expect(actual[key], key).toBe(value);
}
}, 60000);
it("still interpolates a ${VAR} self-reference in the .env file", () => {
mkdirSync(codePath, { recursive: true });
const serviceEnv = [
"APP_URL=https://example.com",
'ASSET_URL="${APP_URL}"',
].join("\n");
const command = getCreateEnvFileCommand({
appName,
composePath: "docker-compose.yml",
env: serviceEnv,
randomize: false,
suffix: "",
serverId: null,
environment: { project: { env: "" }, env: "" },
} as Parameters<typeof getCreateEnvFileCommand>[0]);
execFileSync("bash", ["-c", command]);
const composeFile =
"services:\n test:\n image: busybox\n environment:\n - ASSET_URL=${ASSET_URL}\n";
writeFileSync(join(codePath, "docker-compose.yml"), composeFile);
const out = execFileSync(
"docker",
[
"compose",
"run",
"--rm",
"-T",
"test",
"sh",
"-c",
"printf '%s' \"$ASSET_URL\"",
],
{ cwd: codePath, encoding: "utf8" },
);
expect(out).toBe("https://example.com");
}, 60000);
});
+51
View File
@@ -1,5 +1,6 @@
import {
prepareEnvironmentVariables,
prepareEnvironmentVariablesForFile,
prepareEnvironmentVariablesForShell,
} from "@dokploy/server/index";
import { describe, expect, it } from "vitest";
@@ -642,3 +643,53 @@ SPECIAL=café résumé naïve
expect(resolved[2]).toContain("café");
});
});
describe("prepareEnvironmentVariablesForFile (.env file escaping)", () => {
it("keeps a well-formed ${VAR} self-reference interpolatable", () => {
const serviceEnv = `
APP_URL=https://example.com
ASSET_URL=\${APP_URL}
`;
const resolved = prepareEnvironmentVariablesForFile(serviceEnv, "", "");
expect(resolved).toEqual([
`APP_URL="https://example.com"`,
`ASSET_URL="\${APP_URL}"`,
]);
});
it("keeps a ${VAR:-default} self-reference with a modifier interpolatable", () => {
const serviceEnv = `ASSET_URL=\${APP_URL:-https://default.com}`;
const resolved = prepareEnvironmentVariablesForFile(serviceEnv, "", "");
expect(resolved).toEqual([`ASSET_URL="\${APP_URL:-https://default.com}"`]);
});
it("escapes a literal dollar sign followed by a word so it isn't interpolated", () => {
const serviceEnv = "PASSWORD=pa$word";
const resolved = prepareEnvironmentVariablesForFile(serviceEnv, "", "");
expect(resolved).toEqual([`PASSWORD="pa\\$word"`]);
});
it("escapes a trailing dollar sign", () => {
const serviceEnv = "PRICE=100$";
const resolved = prepareEnvironmentVariablesForFile(serviceEnv, "", "");
expect(resolved).toEqual([`PRICE="100\\$"`]);
});
it("still escapes double quotes and backslashes", () => {
const serviceEnv = String.raw`MESSAGE=say "hi" and \backslash`;
const resolved = prepareEnvironmentVariablesForFile(serviceEnv, "", "");
expect(resolved).toEqual([
String.raw`MESSAGE="say \"hi\" and \\backslash"`,
]);
});
});
@@ -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 = `${SERVICE_NAME}.1.qh3ldvg2h9x0test`;
const task = await docker.createContainer({
name: taskName,
Image: TEST_IMAGE,
Cmd: ["sleep", "3600"],
});
await task.start();
await createLegacyZombie();
await setupMonitoring("test-server");
expect(await containerExists(SERVICE_NAME)).toBe(false);
expect(await containerExists(taskName)).toBe(true);
const inspect = await docker.getContainer(taskName).inspect();
expect(inspect.State.Running).toBe(true);
},
REAL_TEST_TIMEOUT,
);
it(
"is idempotent when no legacy container exists",
async () => {
expect(await containerExists(SERVICE_NAME)).toBe(false);
await expect(setupMonitoring("test-server")).resolves.not.toThrow();
await expect(setupMonitoring("test-server")).resolves.not.toThrow();
expect(await serviceExists(SERVICE_NAME)).toBe(true);
},
REAL_TEST_TIMEOUT,
);
it(
"deploys the service even when removing the legacy container fails",
async () => {
const failingDocker = {
getContainer: () => ({
remove: async () => {
const error: any = new Error("device or resource busy");
error.statusCode = 500;
throw error;
},
}),
getService: docker.getService.bind(docker),
createService: docker.createService.bind(docker),
};
const remoteDocker = await import(
"@dokploy/server/utils/servers/remote-docker"
);
const spy = vi
.spyOn(remoteDocker, "getRemoteDocker")
.mockResolvedValue(failingDocker as any);
try {
await expect(setupMonitoring("test-server")).resolves.not.toThrow();
expect(spy).toHaveBeenCalled(); // guards against the spy silently not intercepting
expect(await serviceExists(SERVICE_NAME)).toBe(true);
} finally {
spy.mockRestore();
}
},
REAL_TEST_TIMEOUT,
);
},
);
@@ -94,4 +94,32 @@ describe("readValidDirectory (path traversal)", () => {
),
).toBe(true);
});
it("returns true for SvelteKit routes with + prefix and @ symbols", () => {
expect(
readValidDirectory(
`${BASE}/applications/myapp/code/src/routes/+page.svelte`,
),
).toBe(true);
expect(
readValidDirectory(
`${BASE}/applications/myapp/code/src/routes/+layout.svelte`,
),
).toBe(true);
expect(
readValidDirectory(
`${BASE}/applications/myapp/code/src/routes/+server.ts`,
),
).toBe(true);
expect(
readValidDirectory(
`${BASE}/applications/myapp/code/src/routes/+error.svelte`,
),
).toBe(true);
expect(
readValidDirectory(
`${BASE}/applications/myapp/code/node_modules/@types/node/index.d.ts`,
),
).toBe(true);
});
});
@@ -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: ${error?.message ?? error}`,
);
}
}
};
const deployMonitoringService = async (
docker: Awaited<ReturnType<typeof getRemoteDocker>>,
serviceName: string,
settings: CreateServiceOptions,
) => {
await removeLegacyContainer(docker, serviceName);
try {
const service = docker.getService(serviceName);
const inspect = await service.inspect();
@@ -86,6 +86,7 @@ export const serverSetup = async (
...server.metricsConfig.server,
token: token,
urlCallback: urlCallback,
cronJob: server.metricsConfig.server.cronJob || "0 0 * * *",
},
containers: server.metricsConfig.containers,
},
+1 -1
View File
@@ -548,7 +548,7 @@ export const prepareEnvironmentVariablesForFile = (
const escapedValue = value
.replace(/\\/g, "\\\\")
.replace(/"/g, '\\"')
.replace(/\$/g, "\\$");
.replace(/\$(?!\{[A-Za-z_][A-Za-z0-9_]*(?::?[-+?][^}]*)?\})/g, "\\$");
return `${key}="${escapedValue}"`;
});
};
+1 -1
View File
@@ -40,7 +40,7 @@ export const readValidDirectory = (
directory: string,
serverId?: string | null,
) => {
if (!/^[\w/. :[\]-]{1,500}$/.test(directory)) {
if (!/^[\w/. :[\]+@~(),=%-]{1,500}$/.test(directory)) {
return false;
}
+4 -1
View File
@@ -18,6 +18,9 @@ async function findFreePort(start) {
return port;
}
const start = Number.parseInt(process.argv[2] || process.env.PORT || "3000", 10);
const start = Number.parseInt(
process.argv[2] || process.env.PORT || "3000",
10,
);
const port = await findFreePort(start);
process.stdout.write(String(port));
+17
View File
@@ -0,0 +1,17 @@
#!/usr/bin/env bash
set -euo pipefail
PAYLOAD=$(cat)
WORKTREE_PATH=$(echo "$PAYLOAD" | jq -r '.tool_response.worktreePath // empty')
if [ -z "$WORKTREE_PATH" ]; then
exit 0
fi
# Turbopack refuses to resolve through anything outside its detected
# workspace root, so a symlinked node_modules (whole dir or per-entry)
# doesn't work for apps/dokploy. A real `pnpm install` is required, but
# since pnpm's global content-addressable store is already warm, this
# only links locally — no network fetch, a few seconds.
cd "$WORKTREE_PATH"
pnpm install --prefer-offline
+55
View File
@@ -0,0 +1,55 @@
#!/usr/bin/env bash
set -euo pipefail
NAME="${1:?usage: spawn-agent-worktree.sh <name>}"
# --show-toplevel would return the CURRENT worktree's own root if this is
# run from inside one (e.g. another agent's worktree) instead of the main
# checkout. --git-common-dir always points at the shared .git regardless of
# which worktree you're standing in.
REPO_ROOT="$(dirname "$(git rev-parse --path-format=absolute --git-common-dir)")"
WORKTREE_PATH="$REPO_ROOT/.claude/worktrees/$NAME"
BRANCH="worktree-$NAME"
if [ -e "$WORKTREE_PATH" ]; then
echo "Worktree already exists: $WORKTREE_PATH" >&2
exit 1
fi
git -C "$REPO_ROOT" fetch origin canary --quiet
git -C "$REPO_ROOT" worktree add -b "$BRANCH" "$WORKTREE_PATH" origin/canary >&2
# .worktreeinclude lists gitignored files (.env, .env.local) that a plain
# `git worktree add` won't check out on its own - copy them in.
while IFS= read -r pattern; do
[ -z "$pattern" ] && continue
find "$REPO_ROOT" \( -path "$REPO_ROOT/.claude/worktrees" -o -path "$REPO_ROOT/node_modules" \) -prune -o -name "$pattern" -print 2>/dev/null
done < "$REPO_ROOT/.worktreeinclude" | while read -r src; do
rel="${src#"$REPO_ROOT"/}"
dest="$WORKTREE_PATH/$rel"
mkdir -p "$(dirname "$dest")"
cp "$src" "$dest"
done
cd "$WORKTREE_PATH"
pnpm install --prefer-offline >&2
FREE_PORT=$(node "$REPO_ROOT/scripts/find-free-port.mjs")
sed -i.bak "s/^PORT=.*/PORT=$FREE_PORT/" apps/dokploy/.env
sed -i.bak -E "s#^(BETTER_AUTH_URL=https?://[^:/]+):[0-9]+#\1:$FREE_PORT#" apps/dokploy/.env
rm -f apps/dokploy/.env.bak
pnpm --filter=dokploy run dev > "$WORKTREE_PATH/dev-server.log" 2>&1 &
echo $! > "$WORKTREE_PATH/dev-server.pid"
BASE_URL="http://localhost:$FREE_PORT"
for _ in $(seq 1 30); do
CODE=$(curl -s -o /dev/null -w '%{http_code}' --max-time 1 "$BASE_URL/" || true)
if [ "$CODE" != "000" ]; then
break
fi
sleep 1
done
echo "export WORKTREE_PATH=$WORKTREE_PATH"
echo "export DOKPLOY_BASE_URL=$BASE_URL"
echo "export PORT=$FREE_PORT"