Documents the gotcha where convert_hf_to_gguf.py crashes with
'config.json not found' because the training output directory holds
only the LoRA adapter, not a merged HF model. Includes inline
save_pretrained_merged() fix snippet, verification checklist, and
resume-pipeline-without-retraining pattern.
Discovered today during the MajorTwin v8c pipeline failure (Step 4).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Documents the non-obvious failure mode where /etc/hosts generator scripts
using `tailscale status --json | jq '.HostName'` get poisoned by iOS
peers, which always report HostName as the literal string "localhost"
because iOS doesn't expose the device name to apps.
Includes the buggy and fixed jq filter (use .DNSName first label
instead), a real-world Postfix outage example, and a verification
checklist. Linked from troubleshooting index and SUMMARY.
Discovered while diagnosing a 24h Postfix outage on majordiscord.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two articles surfaced during the v8 deploy + eval on 2026-04-25:
- Ollama: `ollama run` with piped stdin bypasses the chat template and
SYSTEM prompt — output looks like raw base-model completion. Caught
during initial v8 smoke test. Fix: use /api/chat HTTP endpoint.
- rsync over Tailscale can hang in TCP teardown after the data has
fully transferred. Verify with md5sum, then kill the hung pipeline.
Includes a watcher-threshold gotcha (set below true file size, not
above) and prevention tips.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Documents why `ansible myhost -m ping` fails with Permission denied
while `ssh myhost` works — SSH Host blocks match on literal pattern,
not on resolved HostName, so `ansible_host: <IP>` bypasses the alias
and the declared IdentityFile never gets applied. Covers the portable
fix (ansible_ssh_private_key_file in host_vars), the symlink sidebar
for standardizing key names across control nodes, alternatives, and
a diagnosis checklist.
Also catches index.md up with the ansible-check-mode-false-positives
article that was already published but missing from the nav.
Documents the cosmetic but persistent warning during dnf upgrades:
"/usr/sbin cannot be merged yet, /usr/sbin/ebtables points to
/etc/alternatives/ebtables"
Stale update-alternatives symlinks (not rpm-owned) block Fedora's
/usr/sbin -> /usr/bin consolidation. Article covers root cause,
investigation steps, and the fix (tear down + re-register with
/usr/bin paths only). References the Ansible playbook
fix_ebtables_usrmerge.yml that implements this fleet-wide.
Applied 2026-04-19 across majorlab, majorhome, majormail, majordiscord.
claude-mem 12.1.3 passes --setting-sources with no value, which Claude Code
2.1.x rejects. Documents the silent summaryStored=null symptom, the real
error revealed under DEBUG logging, and the claude-shim workaround.
Captures the majorlab incident where the backup watchdog emailed a missing
heartbeat after a kernel-update reboot wiped /var/run, even though the
backup had actually completed cleanly. Documents the tmpfs root cause and
the fix of storing heartbeats under /var/lib instead.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Covers shell quoting for URLs containing &, ?, #, and other characters
that Bash interprets as operators. Common gotcha when downloading from
CDNs with token-based URLs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Corrected inflated article count (was 76, actual is 73).
Updated domain breakdown and frontmatter timestamps from Obsidian.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Documents the SSH keepalive fix for dnf upgrade timeouts on Fedora hosts,
plus the do-agent task guard fix. Also adds Ansible & Fleet Management
section to the troubleshooting index.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Documents the 2026-03-14 incident where MajorAir's public IP was banned
by the postfix-sasl jail after repeated SASL auth failures, silently
blocking all IMAP connections from Spark Desktop.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add troubleshooting article covering the three-part failure mode on
Fedora with SELinux Enforcing: docker.socket disabled, ports 4443/8448
blocked, and httpd_can_network_connect off. Update index and SUMMARY.
