If you have just upgraded a machine to a shiny new Linux and now MongoDB will not stay up, the problem may well be the kernel rather than your data or configuration.

I lost an afternoon to this last week, so here is the short version to save you the same.

The symptom⌗

On my machine, mongod would start, run for about sixty seconds, and then just disappear.

That exact timing is my experience, not a promise that every affected setup dies on the same schedule. But the overall pattern is the interesting bit: MongoDB comes up, looks fine for a moment, and then falls over again for no obvious reason.

If you are running it under systemd, Docker, Kubernetes, or anything else with a pulse, you then get the added joy of a restart loop.

What’s actually going on⌗

MongoDB have documented a known incompatibility between MongoDB 8.0 and newer and Linux kernel 6.19 and later. Their public explanation points to the version of TCMalloc currently vendored in MongoDB, not to bad data files and not to some weird one-off local misconfiguration.

That means an OS upgrade can be the thing that breaks a previously healthy mongod, even when nothing about your database itself changed.

So if your first instinct was “what did I do to my data?”, the answer may be: nothing at all.

Why you’re about to see more of this⌗

Ubuntu 26.04 LTS ships with Linux 7.0. Kubuntu 26.04 does too. That matters because 7.0 is firmly in the kernel range that this issue affects.

At the same time, Ubuntu 25.10 reaches end of life on 9 July 2026, which means plenty of people are being nudged onto newer releases right now whether they asked for that excitement or not.

So yes: “I upgraded Ubuntu and now MongoDB keeps crashing” is, unfortunately, where we are. And, importantly, because this is a kernel incompatibility - don’t think because you are running inside containers you are safe.

Getting it working again⌗

There are really only three sane responses.

1. Move back to an older kernel⌗

This is the cleanest short-term fix if you control the host. MongoDB’s current guidance is essentially to get off the affected kernel range and wait for a fixed build.

If you can boot an older kernel and stay there for a bit, that is the least surprising path back to a working database.

2. Stay off MongoDB 8.x for now⌗

If you cannot change the kernel, the practical answer may be to avoid the affected MongoDB major line on that machine.

Do not freestyle this on a production database. Check MongoDB’s compatibility, downgrade, and upgrade guidance before touching real data. If this is just a disposable dev instance, life is less dramatic.

3. Treat clever tunables as experiments, not fixes⌗

You may find per-process workarounds, environment tweaks, and forum folklore that claim to make the crashes stop.

Maybe they do, on some setups. Fine. But verify them; it’s possible that they work on ‘bare’ instances, but I couldn’t find any way that helped my instances inside containers stay up.

The short version⌗

  • MongoDB 8.0 and newer has a documented incompatibility with Linux kernel 6.19 and later.
  • Ubuntu 26.04 and Kubuntu 26.04 ship with Linux 7.0, so ordinary distro upgrades can surface it.
  • On my machine, the symptom was a mongod container that died roughly once a minute.
  • MongoDB’s public explanation points to the currently vendored TCMalloc, not to damaged data.
  • The least exciting fix is also the best one: move back to an older kernel and wait for an upstream fix.

Further reading⌗