When Kubernetes Pods CrashLoop for an Unexpected Reason: The ‘Argument List Too Long’ Saga

A few weeks ago, I was called in to help with a Kubernetes issue that, at first glance, seemed like just another pod health blip. But the error message was oddly specific:

exec /usr/local/bin/wait_for.sh: argument list too long

I’ve seen pods spiral into CrashLoopBackOff for all sorts of things: missing configurations, image pull errors, even typos in commands. But this? Was someone passing a novel into bash, or had we finally hit some sort of container epiphany?

Jokes aside, this message pointed to something more subtle happening under the hood, in this case, the maximum limit Linux imposes on the length of process arguments. Now, Kubernetes typically takes care of setting up your service connections automatically, often by populating containers with a bunch of environment variables. These variables are generally harmless, but in large clusters or setups with tons of services and ports, they can quietly balloon to the point where the combined size becomes a problem.

As it turned out, the combination of many services and their associated environment variables pushed us past the system’s limit. Suddenly, every new container started with an environment so big that basic scripts couldn’t run, hitting this “argument list too long” wall and immediately crashing.

So, what’s the fix? Kubernetes supports two main ways for pods to discover services: environment variable injection and DNS lookup. The out-of-the-box experience is powered by environment variables, but DNS is often more scalable and flexible, especially when you have CoreDNS set up.

To test if too many environment variables were the culprit, I disabled the automatic injection for just two deployments using:

kubectl -n <NAMESPACE> patch deployment <NAME> -p '{"spec":{"template":{"spec":{"enableServiceLinks":false}}}}'

Almost instantly, the pods moved out of CrashLoopBackOff and settled into a healthy running state.

Takeaway:
If you ever see a CrashLoopBackOff with this cryptic “argument list too long” error in Kubernetes, it’s worth checking how many environment variables are being injected. Disabling enableServiceLinks could be the quick, safe solution, especially if you’re already using DNS-based service discovery.

Sometimes, the enemy isn’t a misbehaving app, but the well-intentioned features meant to make life easier. Keep an eye on those environment variables. They multiply faster than you think!

Rescuing references:

KodeKloud Troubleshooting Scenario

Good Old Github Issue