Kernel Scenarios — container OOM, Redis BGSAVE, io_uring EPERM, NUMA and the full path of a busy service
local processingFive troubleshooting stories that cross containers, memory and I/O, each a chain of the kernel tools with the right state preloaded: a container OOM-killed with a small RSS, Redis BGSAVE under a memory limit, io_uring failing with EPERM in a container, a container spanning NUMA nodes, and the full path of a service with ten thousand connections. Search by symptom. Runs in your browser.
Why some problems need more than one tool
What it is
Most production problems in containers do not stay in one layer: an OOM kill involves page cache and the cgroup limit, a BGSAVE spike involves THP and fork, an io_uring error involves the kernel, a sysctl and seccomp, and slow memory access involves NUMA placement. Each scenario here is a short chain of the tools in this section, opened with the state that shows the point, with one line at the end saying what to take away.
When to use it
Use it as a map: start from the symptom you have, follow the steps in order, and use the linked tools on your own numbers.
Gotchas & limits
Each step links to a model, not to your machine. The numbers in the preloaded states are examples; replace them with yours, and confirm on a real host with the read-only commands the tools provide.