rdmsm4x-changelog-20260926-1951-network-unas-raid10-plan-and-fleet-diagnostics

Summary: Executed read-only assessment for UNAS Pro 8 RAID6 to RAID10 migration plan and diagnosed NAS SMB slowness, rdmpw3265m reachability, OrbStack unauthenticated port exposures, and rdmsm4x network-changed churn.

Scope: rdmsm4x (local host, 10GbE en0), unaspro818a (UNAS Pro 8 at 192.168.0.9), rdmpw3275m, rdmbair15m5, rdmpw3265m.
Status: Strictly read-only investigation. No system configurations, storage pools, or files modified outside handoff documentation.
Ticket: ISSUE-20260926-34 commented and claim released.
Fleet Bus: Message 20260926-195045-CF064FB4 sent to network@rdmsm4x on topic network-unas-plan.

1. UNAS Pro 8 RAID6 to RAID10 Migration Plan

Target file: ~/dev/_handoff/network/unas-raid10-plan-20260926.md
Total current capacity: 153 TiB (RAID 6 across 8x 28 TB drives). Total used: 52 TiB.
RAID 6 to RAID 10 in-place conversion is not supported by UniFi Drive; pool deletion is mandatory.
Target RAID 10 provides 101.8 TiB usable (~51% utilization with 52 TiB data).
Active backups (rdmsm4x + rdmbair15m5) account for ~10.4 TB. Stale backups (rdmpw3275m, rdmpw3265m, rdmbairM3, rdmbpM2x) account for ~30.5 TB.
Staging host rdmpw3275m has 18 TiB free, sufficient for active backups but requires external storage if all 52 TiB must be preserved.

2. NAS SMB Slowness RCA

Target file: ~/dev/_handoff/network/findings-20260926.md
Network link: 10Gbase-T full duplex on en0, 0 errors, 0.35ms ping latency, 0% packet loss.
Raw SMB throughput: 402.4 MB/s (3.22 Gbps) sequential read from band files.
Time Machine throughput: 3.48 KB/s (68 KB in 20s, 0.85 files/s).
Cause: Traversal of 26.8M files, client backupd XPC connection invalid loops, and high APFS transaction metadata overhead with synchronous SMB writes under AES_128_GMAC signing.

3. rdmpw3265m Reachability

Tailscale status: offline, last seen 18h ago (~01:21 EDT).
LAN ping to 192.168.1.53 and 192.168.0.177: 100% loss. ARP: no entry. mDNS: absent.
Conclusion: Host is off-network (powered off, in deep sleep, or disconnected).

4. OrbStack Exposed Ports

Ports 3000 (Grafana), 3100 (Loki), 12345 (Alloy UI), and 514 (Syslog UDP/TCP) open on 0.0.0.0 and [::], confirmed reachable from rdmbair15m5.
Grafana has anonymous admin login enabled.
Draft 127.0.0.1 binding prepared for localAI bot in /Users/richh/logging/docker-compose.yml (keeping 514/1514 exposed for network syslog).

5. rdmsm4x "Network Changed" Churn

Identified trigger: configd: AUTOMATIC-V6 en1 firing every 10-15s (~4x/min) due to multiple APs broadcasting IPv6 RAs on Wi-Fi en1 (10.0.4.x) with an ultra-short preferred lifetime of 119 seconds.
Proposed remediation: Disable Wi-Fi en1 on rdmsm4x (desktop on 10GbE) or configure Wi-Fi IPv6 to link-local only.