🦎 Keeping My Pangolin Instance Alive on Oracle Cloud (With a Simple Watchdog)

The Problem

I’m running a Pangolin instance on Oracle Cloud Infrastructure (Free Tier).

Overall? Surprisingly solid.

But every now and then:

• the service stops responding
• the tunnel is dead
• HTTP just hangs

From the outside, it looks like the VM is gone.

But when I check OCI:

👉 The instance is still running

“Doesn’t Oracle detect that?”

Actually… yes and no.

OCI provides an infrastructure metric:

instance_status

• 0 → everything fine
• 1 → infrastructure problem

So Oracle does know when something is wrong at the hypervisor level.

But here’s the catch:

Your application can be completely dead while OCI still says everything is fine.

Existing solutions

There is already a project for this:

👉 https://github.com/afreidah/oracle-watchdog

It’s a solid approach, but for my setup it felt a bit heavy.

I didn’t want:

• additional infrastructure
• complex deployments
• something I don’t fully understand

The idea

Instead of relying purely on OCI, I flipped it around:

👉 Check from the outside — like a real user would

Then combine that with OCI data.

The approach

With the help of AI I built a small Bash script that:

1. Decides what to do

Queries OCI:

instance_status

Checks my public endpoint:

https://pangolin.roksblog.de

The logic

If OCI reports problem → resetElse if HTTP fails multiple times → resetElse → do nothing

Simple. Transparent. No magic.

Why this works better than OCI alone

Because it reflects reality.

OCI sees:

• VM is up
• hypervisor is fine

But your users experience:

• timeouts
• broken service
• no response

👉 The HTTP check is the truth.

Making it safe

The important part is avoiding “panic reboots”.

So the script includes:

🧠 Failure threshold

Only reset after multiple failed checks (e.g. 3)

⏱ Cooldown

Prevent reboot loops

🌐 Smart HTTP handling

Only treat real failures as errors:

• timeout (000)
• 500, 502, 503, 504

Not:

• redirects
• auth errors
• 404s

Bonus: Telegram alerts

Because silent restarts are boring.

The script sends notifications when:

• a failure is detected
• a reset is triggered
• the service recovers

Running it from my homelab

This is the part I like most:

👉 The watchdog runs outside OCI

That means:

• it still works if the VM is completely dead
• no dependency on the instance itself
• no chicken-and-egg problem

Cron job:

*/10 * * * * ~/scripts/oci-watchdog/oci-watchdog.sh

OCI integration

Reset is done via CLI:

oci compute instance action \  --instance-id <INSTANCE_ID> \  --action RESET

Authentication is handled via:

~/.oci/config

Result

Since running this:

• no more “silent hangs”
• automatic recovery
• clear logs and notifications

And most importantly:

👉 I don’t have to think about it anymore

Lessons learned

This is one of those classic cloud lessons:

The provider gives you building blocks — not complete solutions.

OCI gives you:

• metrics
• APIs
• compute

But:
👉 you still need to build the glue

Final thoughts

Could this be more “enterprise”?

Sure:

• OCI Functions
• Monitoring alarms
• full auto-healing pipelines

But honestly?

For a Free Tier setup:

👉 this simple watchdog is perfect

Repo

I’ve put the script and setup here:

👉 https://github.com/r0k5t4r/oci-watchdog

Closing thought

Sometimes the best solution is not:

“more cloud”

but:

“a small script that actually solves the problem”