Skip to content

docs(soc): odin goes on large_data — the SSD pool #527 measured (#266) - #635

Merged
Gerrrt merged 2 commits into
mainfrom
gerrrt/wazuh-saruman-deployment-dc0f83
Sep 22, 2026
Merged

Gerrrt merged 2 commits into
mainfrom
gerrrt/wazuh-saruman-deployment-dc0f83

Conversation

@Gerrrt

@Gerrrt Gerrrt commented Sep 22, 2026

Copy link
Copy Markdown
Owner

Refs #266. Does not close it: #266 closes on the day the six agents report in, and that waits on #414.

Why

All of #266's decisions are made and written in code. The one stale piece was build-the-soc-guest.md, written before #527. It still created odin on local-lvm ("the same spindles as everything else"), which would have left the heaviest writer in ADR-0007 as the only guest on the HDD mirror, even though alexander already lives on large_data.

What changes

  • build-the-soc-guest.md
    • §0's Disk row and §1's qm create now use large_data:96,…,ssd=1. The command is rewritten in place because the page has never been run.
    • Adds a pool-headroom check.
    • Adds the qemu-guest-agent install and its qm guest exec check, which alexander's build found missing on 2026-09-20.
  • ADR-0030: a dated note. The refresh_interval row was priced on ~90 IOPS, and the pool measured 7,952. 30s stays, on the segment-churn half of the argument. Heap-per-shard was always the binding constraint, so the 30-day ISM delete is unchanged.
  • changelog.md: an entry for this change.

Nothing in stacks/soc changes.

Checks

  • make check-docs passed.
  • scripts/lint.sh passed, with shellcheck, actionlint and editorconfig-checker skipped locally because they're unavailable.

🤖 Generated with Claude Code

Gerrrt and others added 2 commits September 22, 2026 15:36
build-the-soc-guest.md predates #527 and still created odin on local-lvm,
leaving the heaviest writer in ADR-0007 as the only guest on the spindles.
§1 now creates it on large_data with ssd=1 (rewritten in place: the page has
never been run) and installs qemu-guest-agent, which alexander's build found
missing. ADR-0030 gets a dated note: refresh_interval stays 30s, on the
segment-churn half of its argument now the IOPS half is gone.

Refs #266

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…deployment-dc0f83

# Conflicts:
#	docs/changelog.md
@Gerrrt
Gerrrt merged commit 9e06fe8 into main Sep 22, 2026
4 checks passed
@Gerrrt
Gerrrt deleted the gerrrt/wazuh-saruman-deployment-dc0f83 branch September 22, 2026 22:47
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