Comparison
EvictionvsLoad shedding
Eviction
a healthy workload was terminated because the node it happened to be on ran out of disk, and it had done nothing wrong.
The platform terminating running workloads to relieve pressure on a node — memory, disk or PIDs — choosing victims by QoS class and by how far over their requests they are. It is a deliberate sacrifice rather than a failure, and it is why a workload can die for reasons entirely outside its own behaviour. Disk pressure from unbounded logs or image sprawl is the most common trigger and the easiest to prevent.
Full entry →Load shedding
you start rejecting requests immediately with a 503 instead of accepting work you have no chance of finishing in time.
Deliberately dropping a fraction of incoming work when the system is beyond capacity, so the remainder is served correctly. Accepting everything and timing out is worse for everyone: the client already gave up, and you spent the capacity anyway. Shed the cheapest-to-reject and least valuable traffic first, which requires knowing which traffic that is before the incident.
Full entry →