azcomputerguru/claudetools
1
0
Fork 0
Code Issues 58 Pull Requests Actions Packages Projects Releases Wiki Activity
Files
c82c1c76bbfbe60b08bb6b08e2288ef830b03859
claudetools/clients/scileppi-law/session-logs/2026-07/2026-07-01-howard-sylvia-server-mount-nosleep-screenconnect.md
Howard Enos e527b89999 sync: auto-sync from HOWARD-HOME at 2026-07-01 11:39:58 
Author: Howard Enos
Machine: HOWARD-HOME
Timestamp: 2026-07-01 11:39:58
2026-07-01 11:41:17 -07:00

11 KiB
Raw Blame History

Scileppi Law — Sylvia "can't connect to server": NAS reboot recovery, durable AFP mount + no-sleep fix, ScreenConnect deploy, ticket #32493

User

  • User: Howard Enos (howard)
  • Machine: Howard-Home
  • Role: tech

Session Summary

Sylvia at The Law Offices of Chris Scileppi reported she "can't connect to the server" — a mapped drive that was no longer mapped. The wiki had the Mac context but the entire server/network layer was undocumented, so the infra was discovered live via GuruRMM. The client's Mac is Mac-mini-2.localdomain (macOS, user sylvia); "mapped drive" = an SMB/AFP mount of the Data share at /Volumes/Data. Auto-mount is driven by a GUI LaunchAgent com.acg.mount-server.

The internal RMM API (172.16.3.30:3001) and coord API (:8001) were both unreachable from Howard-Home (off ACG LAN). Authenticated instead against the public RMM endpoint https://rmm.azcomputerguru.com (same vault creds) and drove everything from there. Both Scileppi agents were online: Mac-mini-2 and SL-SERVER. Root cause: SL-SERVER (a Synology NAS, 192.168.242.5) had rebooted ~09:34 local and was slow to bring SMB/AFP back up. Diagnostics initially showed SMB not listening and no afpd/avahi processes; a few minutes later 139/445/548 were all listening again — the NAS simply took time to start file services after boot. The 25 TB btrfs /volume1 pool (incl. /volume1/Data) was mounted and healthy throughout — no data risk.

Two deeper problems surfaced. (1) The mount LaunchAgent targeted the Bonjour AFP service name afp://SL-SERVER._afpovertcp._tcp.local/Data, which the Synology stops advertising after a reboot (NORESOLVE), while SL-SERVER.local → 192.168.242.5 resolves fine and AFP:548 was open. A manual .local mount as sylvia succeeded immediately. (2) The Mac mini is on Wi-Fi (en1) with displaysleep 10 — when the display slept the Wi-Fi link dropped, which caused the RMM agent to bounce offline repeatedly (compounded by the user cycling Sylvia's login). This "screensaver kills the internet" behavior was the meta-cause of all the churn.

Applied fixes over RMM: pmset -a sleep 0 displaysleep 0 womp 1 powernap 1 tcpkeepalive 1 (system sleep was already 0; the only effective change was displaysleep 10 -> 0, which stopped the network drops); rewrote com.acg.mount-server.plist to use afp://SL-SERVER.local/Data (backed up original, reloaded), and retired the stale duplicate agent com.azcomputerguru.mount-slserver. Installed the ScreenConnect macOS access agent (built a parameterized pkg installer via the screenconnect skill, pushed + installed via installer -pkg -target /); session Mac-mini-2 registered self-tagged Company="The Law Offices of Chris Scileppi" / Site="Main Office" / Tag="Mac". Declined to install ScreenConnect on SL-SERVER (headless Synology; unsupported/no value).

Finally, via the /syncro skill, documented the work on ticket #32493 (id 113284439): a customer-visible resolution comment (issues + fixes, do_not_email: true) and a 0.5h Labor - Remote Business line ($150/hr = $75.00), left un-invoiced per Howard's choice (customer has no prepaid block). Updated the Scileppi wiki with the previously-missing Network/Servers/GuruRMM sections and a dated event.

Key Decisions

  • Drove RMM via the public endpoint https://rmm.azcomputerguru.com because the internal 172.16.3.30:3001 is LAN-only and unreachable from Howard-Home. Same vault creds work.
  • Client-side .local fix over a server-side Bonjour fix. The durable, minimal fix was repointing the Mac's mount to the resolvable SL-SERVER.local rather than chasing why the Synology stopped advertising _afpovertcp._tcp after reboot. .local is robust regardless of AFP Bonjour advertisement.
  • Kept AFP (didn't switch to SMB). AFP:548 works and the Keychain creds are AFP; switching to smb:// would change mount type, permissions, and the stored credential. Noted SMB/AFP-deprecation as a future consideration only.
  • displaysleep 0 (display never sleeps) as the connectivity fix. System sleep was already 0; the display-sleep-triggered Wi-Fi drop was the actual fault. Recommended Ethernet as the bulletproof long-term fix.
  • Grabbed the Mac on first reconnect to slam pmset first, before doing anything else, because the machine kept dropping when idle — pmset is instant, so once applied the machine stayed up for the rest of the work.
  • Did not install ScreenConnect on SL-SERVER — headless Synology DSM is not a supported/useful SC target; it stays managed via GuruRMM + DSM web UI.
  • Added time but did not invoice ticket #32493 — per Howard's explicit choice at the preview (no prepaid block, so an invoice would bill $75).

Problems Encountered

  • Both internal APIs (RMM 3001, coord 8001) timed out from Howard-Home (off-LAN). Resolved by using the public RMM endpoint.
  • Repeated agent drops (undeliverable/interrupted/pending) on Mac-mini-2 mid-command. Root-caused to display-sleep dropping Wi-Fi plus the user cycling Sylvia's login. Resolved by applying displaysleep 0 on first stable reconnect.
  • False-positive mount check (self-inflicted). mount | grep -qi "/Volumes/Data" matched /System/Volumes/Data (the macOS system volume) as a substring and reported "MOUNTED-OK" when the AFP share was absent. Corrected to the precise grep -q "on /Volumes/Data " (trailing space) that the LaunchAgent itself uses. Logged to errorlog.md (--friction).
  • Bonjour AFP name NORESOLVE. The plist's _afpovertcp._tcp.local name didn't resolve post-reboot; SL-SERVER.local did. Fixed the plist.
  • SC install command returned overall exit 1 — cosmetic only: the final ls /Applications/ScreenConnect* glob didn't match the actual app dir name connectwisecontrol-*.app. Install itself succeeded (installer rc=0, daemon + processes present, cloud session registered).

Configuration Changes

  • On Mac-mini-2 (via RMM, not in repo):
    • pmset -a sleep 0 displaysleep 0 womp 1 powernap 1 tcpkeepalive 1 (effective change: displaysleep 10 -> 0).
    • /Users/sylvia/Library/LaunchAgents/com.acg.mount-server.plist — mount URL changed afp://SL-SERVER._afpovertcp._tcp.local/Data -> afp://SL-SERVER.local/Data; reloaded (bootout+bootstrap gui/501). Backup: com.acg.mount-server.plist.bak-20260701.
    • Retired stale duplicate agent: com.azcomputerguru.mount-slserver.plist -> .disabled-20260701 (bootout + rename).
    • Installed ScreenConnect macOS agent: /Applications/connectwisecontrol-1912bf3444b41a08.app + LaunchDaemon connectwisecontrol-1912bf3444b41a08.plist.
  • Repo:
    • wiki/clients/scileppi-law.md — added Network + Servers detail, updated GuruRMM section (enrolled agents, public endpoint note, ScreenConnect), added 2026-07-01 Key Events entry, last_compiled -> 2026-07-01.
    • errorlog.md — one --friction entry (grep /System/Volumes/Data false-match).
  • Syncro ticket #32493: resolution comment (id 421638088) + line item (id 43099595).

Credentials & Secrets

None created, rotated, or newly discovered. RMM auth used existing vault entry infrastructure/gururmm-server.sops.yaml; ScreenConnect used msp-tools/screenconnect.sops.yaml; Syncro used the baked-in per-user key (howard). Temp files holding the RMM JWT were removed at end of session.

Infrastructure & Servers

  • Scileppi subnet: 192.168.242.0/24; gateway/DNS 192.168.242.1.
  • SL-SERVER — Synology NAS, 192.168.242.5 (bond0). btrfs /volume1 (25 TB pool, /dev/mapper/cachedev_0), share Data at /volume1/Data. Serves AFP (548) + SMB (445/139). Resolves reliably as SL-SERVER.local; the _afpovertcp._tcp.local Bonjour name is unreliable after NAS reboots. netatalk cedarview build; no systemd (Synology init). GuruRMM agent 0186e9d5-e1cc-4603-a81a-adb1f2230702 (os_type reported linux).
  • Mac-mini-2.localdomain — Sylvia's Mac, macOS, user sylvia (uid 501). On Wi-Fi en1 = 192.168.242.154. GuruRMM agent 1386d9fd-ac16-423c-ada0-5abad5b61838. Mount agent com.acg.mount-server (RunAtLoad + StartInterval 300). Has an unused wired Ethernet port (recommended for durable connectivity).
  • GuruRMM: internal API http://172.16.3.30:3001 (LAN-only); public API https://rmm.azcomputerguru.com (works off-LAN, same vault creds). Auth helper .claude/scripts/rmm-auth.sh hardcodes the internal URL — for off-LAN use, auth manually against the public host.
  • ScreenConnect: https://computerguru.screenconnect.com; new session Mac-mini-2, CP1="The Law Offices of Chris Scileppi", CP2="Main Office", CP3="Mac".

Commands & Outputs

  • Off-LAN RMM auth: POST https://rmm.azcomputerguru.com/api/auth/login with vault email/password -> JWT; then GET /api/agents.
  • Mount diagnosis on Mac: dscacheutil -q host -a name SL-SERVER._afpovertcp._tcp.local -> NORESOLVE; SL-SERVER.local -> 192.168.242.5; nc -z -G 3 192.168.242.5 548 -> open.
  • Working manual mount: launchctl asuser 501 sudo -n -u sylvia /usr/bin/osascript -e 'mount volume "afp://SL-SERVER.local/Data"' -> //Sylvia@SL-SERVER.local/Data on /Volumes/Data (afpfs).
  • Precise mount check (use this, not a bare grep): mount | grep -q "on /Volumes/Data ".
  • pmset BEFORE: sleep 0 / displaysleep 10 / womp 1 / powernap 1 / tcpkeepalive 1; AFTER: displaysleep 0.
  • SC installer build: sc.py build-installer --platform pkg --name Mac-mini-2 --company "The Law Offices of Chris Scileppi" --site "Main Office" --tag Mac; install: curl -fsSL -o /tmp/sc.pkg "<url>"; installer -pkg /tmp/sc.pkg -target / -> "upgrade was successful", client processes running.
  • Syncro: ticket #32493 id 113284439; comment 421638088; line item 43099595 (Labor - Remote Business, 0.5, $150, taxable false).

Pending / Incomplete Tasks

  • Verify auto-mount from a clean state. The share was already mounted when the new .local plist loaded, so the kickstart didn't exercise a from-scratch remount via the new config. The .local mount path is proven working manually; a full clean-boot verification is still outstanding (avoided force-unmounting Sylvia's active drive).
  • Move Mac-mini-2 to wired Ethernet — the durable fix for the Wi-Fi idle drops (pmset displaysleep 0 is the interim mitigation). Requires onsite.
  • Ticket #32493 not invoiced — 0.5h remote ($75) sits on the ticket for the normal billing run. Customer has no prepaid block.
  • Optional: if desired, re-post the resolution comment with do_not_email: false to email the client (currently suppressed).
  • rmm-auth.sh enhancement idea: fall back to the public RMM host when the internal IP is unreachable (would have saved the manual auth this session).

Reference Information

  • Syncro ticket: #32493 (id 113284439) — https://computerguru.syncromsp.com/tickets/113284439 ; customer The Law Offices of Chris Scileppi (customer_id 9601863, Syncro).
  • GuruRMM agents: Mac 1386d9fd-ac16-423c-ada0-5abad5b61838; SL-SERVER 0186e9d5-e1cc-4603-a81a-adb1f2230702.
  • ScreenConnect: app connectwisecontrol-1912bf3444b41a08; extension 2d558935-686a-4bd0-9991-07539f5fe749.
  • Skills used: /rmm, screenconnect, /syncro.
  • Wiki: wiki/clients/scileppi-law.md (updated this session).
  • Bot alerts posted: #dev-alerts (RMM remediation summary), #bot-alerts (Syncro #32493).
Reference in New Issue View Git Blame Copy Permalink