Cloud & Containers guides

Trace the boundary that actually failed

Cloud memory observations can describe different scopes: a physical host, a virtual machine, a container group, or one process. A useful investigation begins by identifying which scope constrained the failing workload and which observations belong to the same time interval.

Start with the Cloud Mem Swap topic guide, then read the detailed host-and-container article below. It connects failure evidence with effective limits, swap allowance, backing storage, and infrastructure ownership. The objective is a diagnosis that survives instance replacement rather than a manual change that works once.

Include application-level alternatives in the test plan. Bounded concurrency, queue limits, or smaller supported batches may address the source of demand more directly than an additional swap area. Compare completed useful work, including failures and retries, rather than relying on process survival alone.

The RAM capacity guide helps define acceptable delay and storage headroom. The Linux guide provides system-level context. When a managed service does not expose the necessary host information, record that limitation and use its documented controls rather than assuming the missing layer is unlimited.