Tyk pump not protected against OOM

[monrax]

If memory usage by one or more pump backends increases faster than it is freed, the Tyk Pump process can reach an OOM state that could take down the entire analytics purger with it.

If a susceptible backend goes into memory runaway, the process might exit with error code 137:

Is this expected behavior?

I suppose this can be addressed with the appropriate restart policy when using the Tyk pump container, but I was wondering if perhaps the offending backends should be limited instead of potentially bringing the pump down every time.

[mativm02]

Hey @monrax! Sorry for the late reply.

This is not an expected behavior. If you increase the purge_delay value, the issue is still present?