Hi All,

Currently, we can configure a threshold for the total RAM consumed by all the running containers.

It's good, that limitation keep the router robust itself when facing OOM problems caused by an arbitrary container.

But in a multiple containers setup, It would be better to implement a per-container level MEM limit(aka: like docker). So that we could configure a reasonable threshold for each container independently.
For instance, If 2 of 6 containers have potential memory leak problems. Independent limits could let the kernel kill thoes containers at early stage, not until on of them eat up the memory and reaches the total threshold.

Is this setting not enough?

Code: Select all

container/config/set ram-high=

OP addressed that in the first paragraph; it sets a single aggregate limit for all containers. OP wants a per-container limit. It is one of many reasons calling container.npk “docker” misleads more than helps.

The only path forward is to file a formal request for the feature and hope they agree your use case is worth supporting.

We have added possibility to write swap to external drive, in some cases this may help with OOM