Kubernetes integration detection fails

[faridco]

Description

I get the following in the log

ERROR: Found 2 Headscale processes: 23, 6
ERROR: Integration Kubernetes (k8s) is not available

but process 6 is no related to the headscale according to process list

/app # ps ax
PID   USER     TIME  COMMAND
    1 65535     0:00 /pause
    6 root      0:02 node ./build/server/index.js
   23 root      0:00 /ko-app/headscale serve

I guess the issue somewhere here

headplane/app/server/config/integration/kubernetes.ts

Lines 161 to 167 in 0866ced

	const path = join('/proc', dir, 'cmdline');
	try {
	log.debug('config', 'Reading %s', path);
	const data = await readFile(path, 'utf8');
	if (!data.includes('headscale') && !data.includes('serve')) {
	throw new Error('Found PID without Headscale serve command');
	}

Might be the issue is in C string (0 byte in the end), the cmdline is an array of zero-terminated strings

/app # xxd < /proc/6/cmdline
00000000: 6e6f 6465 002e 2f62 7569 6c64 2f73 6572  node../build/ser
00000010: 7665 722f 696e 6465 782e 6a73 00         ver/index.js.

Headplane Version

0.6.0

Headscale Version

0.26.0

[tale]

Thanks for raising this, let me look into it.

[tale]

Does this work in the next branch? (ghcr.io/tale/headplane:next)

[faridco]

Image tag change led me to the error.

headplane 2025-06-24T06:57:26.798Z [server] INFO: Running Node.js 22.16.0
headplane 2025-06-24T06:57:26.859Z [config] INFO: Found a valid configuration file at /etc/headplane/config.yaml
headplane 2025-06-24T06:57:26.888Z [config] INFO: Loading a .env file (if available)
headplane /app/node_modules/.pnpm/libsql@0.5.13/node_modules/libsql/index.js:93
headplane       this.db = databaseOpen(path, authToken, encryptionCipher, encryptionKey, timeout);
headplane                 ^
headplane
headplane Error: ConnectionFailed("Unable to open connection to local database /var/lib/headplane/hp_persist.db: 14")

it would be nice if version change could migrate to prevent such errors

[tale]

The next branch is a pre-release, it isn't finished yet. And that error is due to a lack of permissions since I decided to change the container to run as nonroot, it's not really migratable, because your volume mount on the host is expecting root permissions and the container runs with the UID 65532. I had to run chown -R 65532:65532 <host_path> to fix it.

Try that for now, test the Kubernetes integration, I'm going to make a decision on preserving root or not before releasing.

[xnilsit]

Using next, it only detects the correct process now.

PID   USER     TIME  COMMAND
    1 65535     0:00 /pause
   21 root      0:03 /ko-app/headscale serve
   37 root      0:06 /nodejs/bin/node /app/build/server/index.js

2025-06-30T11:28:18.259Z [config] DEBUG: Reading /proc/1/comm
2025-06-30T11:28:18.259Z [config] DEBUG: Reading /proc/21/comm
2025-06-30T11:28:18.259Z [config] DEBUG: Reading /proc/37/comm
2025-06-30T11:28:18.260Z [config] DEBUG: Failed to read /proc/1/comm: Error: Found PID with unexpected command: pause
    at file:///app/build/server/assets/index-D7iQNWJ_.js:2897:19
    at async Promise.allSettled (index 0)
    at async KubernetesIntegration.isAvailable (file:///app/build/server/assets/index-D7iQNWJ_.js:2906:23)
    at async loadIntegration (file:///app/build/server/assets/index-D7iQNWJ_.js:3103:17)
    at async file:///app/build/server/assets/index-D7iQNWJ_.js:4886:16
2025-06-30T11:28:18.261Z [config] DEBUG: Failed to read /proc/37/comm: Error: Found PID with unexpected command: node
    at file:///app/build/server/assets/index-D7iQNWJ_.js:2897:19
    at async Promise.allSettled (index 2)
    at async KubernetesIntegration.isAvailable (file:///app/build/server/assets/index-D7iQNWJ_.js:2906:23)
    at async loadIntegration (file:///app/build/server/assets/index-D7iQNWJ_.js:3103:17)
    at async file:///app/build/server/assets/index-D7iQNWJ_.js:4886:16
2025-06-30T11:28:18.261Z [config] DEBUG: Found Headscale processes: [ 21, [length]: 1 ]
2025-06-30T11:28:18.261Z [config] INFO: Found Headscale process with PID: 21

[tale]

So it works?

[NekitoSP]

I faced this problem, and I can confirm it now works! Thank you!

[vizv]

It seems 2f31617 fixed the issue, waiting for release.

I'm currently using the following patch as a workaround:

diff --git a/base/deployment.yaml b/base/deployment.yaml
index 9636797..a73d39e 100644
--- a/base/deployment.yaml
+++ b/base/deployment.yaml
@@ -37,6 +37,12 @@ spec:
             port: http-headscale
       - name: headplane
         image: ghcr.io/tale/headplane:0.6.0
+        command: [/bin/sh, -c]
+        args:
+        - |
+          sed -i 's#data\.includes("serve")#data.split("\\0").includes("serve")#g' \
+                 ./build/server/assets/index-DPXTyKib.js \
+              && exec node ./build/server/index.js
         env:
         - name: HEADPLANE_LOAD_ENV_OVERRIDES
           value: 'true'