ops: Disk-full / eviction signal on metrics + logs #36

New Issue

2026-09-04T13:51:03-05:00

eva commented

2026-09-04 13:51:03 -05:00

Problem

When the disk tier is full or thrashing eviction, ops look like “random misses” with no clear “we’re dropping data.”

Why

Operators need to distinguish cold cache from capacity pressure.

Scope

One /metrics counter + log when eviction or ENOSPC bites
Tie to the disk-ready signal from #33 (same PR or tiny follow-up OK)

Acceptance

Eviction / ENOSPC visible on /metrics and logs
Documented next to #33 attach-ready notes

Out of scope

#31 HEAD; redesigning GC algorithms

Size: S. From Forge Meeting (Ash). Pair with #33 / Pike occupancy.

## Problem When the disk tier is full or thrashing eviction, ops look like “random misses” with no clear “we’re dropping data.” ## Why Operators need to distinguish cold cache from capacity pressure. ## Scope - One `/metrics` counter + log when eviction or ENOSPC bites - Tie to the disk-ready signal from #33 (same PR or tiny follow-up OK) ## Acceptance - Eviction / ENOSPC visible on `/metrics` and logs - Documented next to #33 attach-ready notes ## Out of scope - #31 HEAD; redesigning GC algorithms Size: S. From Forge Meeting (Ash). Pair with #33 / Pike occupancy.

s1d3sw1ped was assigned by eva

2026-09-04 13:51:03 -05:00

s1d3sw1ped was unassigned by eva

2026-09-04 15:09:07 -05:00

ash was assigned by eva

2026-09-04 15:09:07 -05:00

ash referenced this issue from a commit

2026-09-07 14:23:57 -05:00

ops: Signal disk-full and eviction capacity pressure

ash referenced a pull request that will close this issue

2026-09-07 14:24:57 -05:00

ops: Signal disk-full and eviction capacity pressure #46