Memory allocation in the QM partition

Memory allocation in the QM partition¶

Memory management is crucial to the smooth performance, stability, and protection of an operating system. Efficient allocation of memory supports process isolation and the ability of an operating system to run multiple processes at once. Memory allocation prevents system crashes and performance slowdowns.

The memory_limit attributes max and high control how much memory is allocated to the QM partition.

max controls the absolute limit for memory use in the QM partition. You can set the max attribute in bytes or as a percentage value. The default value of max is infinity, which permits the system to consume as much memory as the system will allow.
high controls the throttling limit for memory use in the QM partition. The value of high can also be set in bytes or as a percentage value. The default value of high is infinity, which essentially eliminates process throttling in times of high memory use.

You can configure memory_limit values in the qm stage of your Automotive Image Builder manifest file to control memory allocation.

```console title="Example memory allocation values" qm: memory_limit: max: 50% high: 45%


## Out of Memory killer

The Linux kernel memory management system includes the Out of Memory (OOM) killer. The OOM killer terminates processes to release memory when
memory becomes critically low. This mechanism ensures that critical processes in the root partition have enough memory to continue to run.

The `oom_score_adj` parameter influences the likelihood that a process will be terminated by the OOM killer. When an operating system experiences
memory constraints, the OOM killer uses the processes' `oom_score_adj` values to prioritize which processes to terminate.

The value of `oom_score_adj` can range from `1000` to `-1000`. The OOM killer terminates processes with the highest `oom_score_adj`
values first. If a process is set to `-1000`, the OOM killer cannot terminate it, whereas the OOM killer would terminate a process set to
`1000` first.

When Podman starts a container nested in the QM partition, it writes the default value of `750` into `/proc/$PID/oom_score_adj`. For more
information about `oom_score_adj`, refer to `man 5 proc_pid_oom_score_adj`.

## OOM protection in Automotive Image Builder

Automotive Image Builder provides an `enable_oom_protection` option that sets
`DefaultOOMScoreAdjust=-1000` in the systemd configuration. When enabled, all host-level
processes receive the lowest possible OOM score by default, which protects them from OOM
termination.

When the QM partition is active (that is, the `qm` section is present in the build manifest), OOM protection
is enabled automatically.

You can also enable or disable OOM protection explicitly in
the `image` section of the manifest by setting `enable_oom_protection` to `true` or `false`, as shown in the following example:

```yaml title="Example enable_oom_protection configuration"
image:
  enable_oom_protection: false

For the procedure to configure memory allocation in the QM partition, see Configuring memory allocation.