Skip to content

feat: containerd on state volume, DLM simplification, unified snapshot queue#7

Merged
brfid merged 4 commits intomainfrom
chore/uptime-estimate
Feb 23, 2026
Merged

feat: containerd on state volume, DLM simplification, unified snapshot queue#7
brfid merged 4 commits intomainfrom
chore/uptime-estimate

Conversation

@brfid
Copy link
Copy Markdown
Owner

@brfid brfid commented Feb 23, 2026

Summary

  • Move containerd data-root to state volume (/opt/edcloud/state/containerd) so Docker image layers survive reprovision; grow both EBS defaults to 30 GB
  • Simplify DLM policy to daily/weekly/monthly (drop quarterly tier)
  • Replace dual snapshot systems with a unified CLI-managed queue: single flat pool capped at 3, enforced by prune(3) → snapshot → prune(3) on every trigger; DLM auto-wiring removed from edc provision/edc destroy

Snapshot triggers

Command Description Blocking
edc up on-start No
edc provision pre-provision Yes
edc reprovision pre-reprovision Yes
edc destroy pre-destroy Yes

Test plan

  • 166 tests passing locally
  • edc upedc snapshot --list shows ≤3 snapshots, newest tagged on-start
  • edc reprovision → snapshot count stays at 3
  • Manually create extra snapshot → next edc up prunes back to 3
  • edc status → shows snapshot count, no DLM warning

🤖 Generated with Claude Code

brfid and others added 4 commits February 22, 2026 17:08
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Redirect containerd's storage root to /opt/edcloud/state/containerd so
Docker image layers and container data survive reprovision alongside the
rest of the state volume.  Grow both EBS volume defaults to 30 GB to
accommodate the combined footprint (root: OS + dev tools only,
state: home + Docker + containerd data).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Remove the quarterly schedule from the DLM backup policy.  With the
CLI-managed snapshot queue as the primary safety net, the quarterly tier
adds cost and complexity without meaningful benefit for a personal lab.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace dual snapshot systems (DLM auto-wired + CLI manual) with a single
flat pool managed entirely by the CLI.  Each trigger runs prune(3) →
snapshot → prune(3), so the cap enforces itself and drift from a failed
prune heals on the next trigger (worst-case +1).

Snapshot triggers:
- edc up:          on-start, fire-and-forget
- edc provision:   pre-provision, blocking (cleanup path)
- edc reprovision: pre-reprovision, blocking
- edc destroy:     pre-destroy, blocking

DLM auto-wiring removed from edc provision and edc destroy; backup-policy
commands remain available for opt-in DLM on top.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
@brfid brfid merged commit 54cd971 into main Feb 23, 2026
1 check passed
@brfid brfid deleted the chore/uptime-estimate branch February 23, 2026 02:30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant