Skip to content

ops: system baseline script + operational patterns (2026-10-08) - #102

Merged
elasticdotventures merged 3 commits into
mainfrom
ops/system-baseline
Oct 15, 2026
Merged

elasticdotventures merged 3 commits into
mainfrom
ops/system-baseline

Conversation

@elasticdotventures

Copy link
Copy Markdown
Member

What

  • scripts/system-baseline.sh / just ops-baseline: read-only (no sudo) host snapshot as markdown. Host addresses are masked.
  • docs/ops/SYSTEM-BASELINE-2026-10-08.md: the first snapshot.
  • docs/ops/OPS-PATTERNS.md: eight failure patterns from the 2026-10-07 stall and today's unit clean-up. Observed values and proposed (untuned) thresholds are kept separate.

Host changes made alongside this PR (not in the diff)

Unit Cause Action
buildkit ExecStart pointed at a missing path, restarting every 2 s (36,000 times) path corrected to /usr/local/bin/containerd-rootless-setuptool.sh; running, 0 restarts
b00t-hive-serena kubectl has no context, so apply fails every 30 s (2,646 times) disabled; a k0s cluster does exist, only the kubeconfig is missing
b00t-historian, container-spire-agent stable for 18 h; counters were stale reset-failed only

Still needs a human

  1. Kernel log (needs adm group): sudo grep -iE "out of memory|oom-kill|hung task|blocked for more|Xid" /var/log/kern.log.1 /var/log/kern.log
  2. Re-enable serena against k0s:
    sudo k0s kubeconfig admin | install -m600 /dev/stdin ~/.kube/k0s.yaml
    systemctl --user edit b00t-hive-serena   # add: [Service] / Environment=KUBECONFIG=%h/.kube/k0s.yaml
    systemctl --user enable --now b00t-hive-serena
    
  3. 12 GB of /tmp (RAM) is a cargo target/ in an ended session's scratchpad: /tmp/claude-1000/-home-brianh-promptexecution/70b8d713-e04c-48da-a7d6-161f47b23c26/scratchpad/b00t-pr/target. IO pressure on the spinning root disk is ~77–88% while it exists; whether removing it fixes that is not yet proven.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GgXQRGsprSZc2oXGNj9n8X

elasticdotventures and others added 3 commits October 8, 2026 07:28
scripts/system-baseline.sh (just ops-baseline) captures memory/swap, PSI, tmpfs,
disks, GPU, listeners, running and failing user units and containers as markdown.
docs/ops/OPS-PATTERNS.md consolidates the 2026-10-07 stall and 2026-10-08 unit
clean-up into eight failure patterns, with observed values and proposed (untuned)
thresholds kept separate. The first snapshot is committed beside it.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GgXQRGsprSZc2oXGNj9n8X
… it cannot run

Without adm membership the section states it is unreadable and gives the exact
command, instead of silently reporting a clean log.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GgXQRGsprSZc2oXGNj9n8X
kern.log shows one real global OOM (2026-10-02, k0s pods then llama-server),
containerd hung-task episodes on 09-29 and 10-05, nothing on 10-07, no GPU Xid.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GgXQRGsprSZc2oXGNj9n8X
@elasticdotventures
elasticdotventures marked this pull request as ready for review October 15, 2026 11:27
@elasticdotventures
elasticdotventures merged commit 04887c3 into main Oct 15, 2026
3 checks passed
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