fix(core): increase ephemeral storage overhead to 200M

[LopatinDmitr]

Description

Increase ephemeralStorageOverheadSize from 50M to 200M in the 3p-kubevirt dependency.

This is a temporary hotfix. The root cause — unbounded growth of virt-launcher stdout/stderr logs — is addressed in a separate task.

Related: deckhouse/3p-kubevirt#134

Why do we need it, and what problem does it solve?

When a node runs low on ephemeral-storage, the kubelet evicts pods ordered by excess usage (actual usage − request). The d8v-compute container accumulates stdout/stderr logs over time; once log size exceeds the 50M request, the pod becomes a prime eviction candidate — which causes the VM running inside it to be killed.

Observed in production:

• Eviction event: d8v-compute used ~63M against a 50M request → VM killed.
• Long-running pods show log sizes up to 105M, well above the 50M threshold.

Raising the request to 200M gives significantly more headroom before a virt-launcher pod is selected as an eviction victim, reducing unintended VM downtime caused by log accumulation.

What is the expected result?

1. Deploy updated virt-launcher with new ephemeral-storage request (200M).
2. Confirm d8v-compute container has resources.requests.ephemeral-storage: 200M.
3. Verify virt-launcher pods are no longer evicted under moderate node disk pressure.

Checklist

• The code is covered by unit tests.
• e2e tests passed.
• Documentation updated according to the changes.
• Changes were tested in the Kubernetes cluster manually.

Changelog entries

section: core
type: fix
summary: "Fixed unexpected VM termination under node disk pressure — VMs are now resilient to moderate ephemeral storage consumption."