Repository navigation
Snapshots: once a day, keep 14 - #163
Conversation
COTEL_SNAPSHOT_INTERVAL 6h -> 24h and COTEL_SNAPSHOT_KEEP 56 -> 14, in the binary's defaults and in docker-compose.yml. Reach stays 14 days; a restore now loses up to a day of telemetry instead of six hours, for about a quarter of the disk. The owner's call, recorded as an amendment to ADR-0023. The probe's SNAPSHOT_STALE_AFTER_SECONDS default moves from 43200 to 172800 (two intervals) so a healthy daily worker does not read as stale. The rpi5 wrapper and the loopback job both take the script's default. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
No actionable comments were generated in the recent review. Walkthrough
Merge Risk:
|
Why
The owner wants daily snapshots instead of every six hours. The 6h × 56 cadence was chosen in ADR-0023. This PR records the change as an amendment to that ADR.
What
COTEL_SNAPSHOT_INTERVAL6h → 24h,COTEL_SNAPSHOT_KEEP56 → 14, ininternal/storage/snapshot.godefaults and indocker-compose.yml. Reach stays 14 days. A restore now loses up to a day of telemetry (was 6h). Disk drops from ~235 MB to ~60 MB.scripts/probe-healthz.sh: theSNAPSHOT_STALE_AFTER_SECONDSdefault goes 43200 → 172800 (two intervals). Without it, a healthy daily worker would page every day. The rpi5 wrapper materializes the script frommainand the loopback job uses the script's default, so neither needs an edit.Rollout note
Pruning keeps the N newest snapshots. On the first daily export, the 6-hourly backlog on prod would shrink to its 14 newest, about 3 days of reach. Thinning the backlog to one snapshot per UTC day before then keeps the full 14 days.
Verification
scripts/probe-healthz_test.sh: 31/31 pass.internal/storagepass.🤖 Generated with Claude Code
Summary by CodeRabbit