UPES-ECS × Juniper — Live Network Discovery (read-only)¶
What this is. A snapshot of what is actually on the wire right now, extracted read-only and without credentials from the PBX laptop. No device configuration was changed and nothing was logged into. Companion to the planning docs (Juniper-Integration-Plan.md, Juniper.md, Juniper-Feature-Catalogue.md) — those say what we could build; this says what is present today.
Captured: 2026-07-18 · From: laptop 192.168.1.6 (Realtek GbE Ethernet, MAC 04:7C:16:A8:06:FA) · Method: ICMP/ARP sweep, TCP port probing, SSH/TLS banner grab, SNMP/NTP/mDNS queries — all unauthenticated. Scope: the one subnet the laptop's wired link is on, 192.168.1.0/24.
1. Executive summary¶
- The Juniper kit is online and healthy on a single flat
192.168.1.0/24— this is the "flat pilot" stage (Juniper.md), not the VLAN/Virtual-Chassis architecture from the plan. So we are effectively at Phase 0. - Confirmed devices: 1× Juniper SRX (gateway/firewall/DHCP/NTP), 2× Juniper EX2300-C switches, 1× HP ProDesk 600 G4 desktop (admin PC), and 1 unidentified endpoint. Plus this PBX laptop.
- The 2× Mist AP32s are NOT on this segment yet (no ARP presence) — they appear undeployed/unpowered here.
- Management is wide open to us (read-only-capable): NETCONF (830) + SSH (22) on all three Junos boxes, J-Web (443) on the SRX. This is exactly what the plan's app-integration ideas (caller-location, network-health panel, evac-mode) need — the transport is ready today.
- Good security posture: SNMP
publicis off; J-Web is TLS 1.2/AES256. To harden: telnet (23) is open on both switches; the network is flat (no segmentation yet).
2. Topology (as discovered)¶
Internet / uplink (via SRX; WARP VPN active on the laptop)
|
┌──────────┴───────────┐
│ Juniper SRX │ 192.168.1.1 (A4:7F:1B:96:85:3B)
│ gateway · DHCP · NTP │ J-Web, NETCONF, SSH
│ serial CV3225AX0113 │ OpenSSH_7.5 (older Junos)
└──────────┬───────────┘
│ (flat L2 — one subnet, no VLANs)
┌──────────────┬──────────┴─────┬───────────────┬─────────────────┐
│ │ │ │ │
┌────┴────┐ ┌─────┴────┐ ┌─────┴────┐ ┌──────┴─────┐ ┌──────┴──────┐
│ EX2300-C │ │ EX2300-C │ │ HP ProDesk│ │ endpoint │ │ PBX laptop │
│ #1 .3 │ │ #2 .4 │ │ 600 G4 .2 │ │ .5 (?) │ │ .6 (this) │
│ E8:A5:5A │ │ E8:A5:5A │ │ F4:39:09 │ │ 7C:B6:8D │ │ 04:7C:16 │
│ SSH/telnet│ │ SSH/telnet│ │ admin PC │ │ unknown │ │ Realtek GbE │
│ NETCONF │ │ NETCONF │ │ (firewalled)│ │ no ports │ │ │
│ OpenSSH_9.7│ │ OpenSSH_9.7│ └───────────┘ └────────────┘ └─────────────┘
└───────────┘ └───────────┘
(identical units; consecutive MACs .9A:77 / .A4:72)
NOT SEEN on this segment: 2× Mist AP32 (undeployed / unpowered here)
Wiring caveat: whether
.3/.4are cabled to each other, to the SRX, and which physical ports the endpoints use is not knowable without reading the switch (LLDP / MAC table). The diagram shows L3 adjacency (all on one subnet), not the physical cabling. See §7 to close this.
3. Device inventory¶
| IP | MAC (OUI) | Identity | Confidence | Open TCP | Fingerprint evidence |
|---|---|---|---|---|---|
| 192.168.1.1 | A4:7F:1B:96:85:3B |
Juniper SRX — gateway, firewall, DHCP server, NTP server | Confirmed | 22, 443, 830 | J-Web <title>Juniper Web Device Manager, CSP → download.juniper.net; TLS cert CN=CV3225AX0113 (serial); SSH-2.0-OpenSSH_7.5; is our DHCP + NTP server |
| 192.168.1.3 | E8:A5:5A:28:9A:77 (Juniper) |
Juniper EX2300-C switch #1 | Confirmed | 22, 23, 830 | Juniper OUI; NETCONF 830; SSH-2.0-OpenSSH_9.7 with CVE-2024-6387/39894 fixes (modern, patched Junos) |
| 192.168.1.4 | E8:A5:5A:28:A4:72 (Juniper) |
Juniper EX2300-C switch #2 | Confirmed | 22, 23, 830 | Identical banner + sibling MAC to #1 |
| 192.168.1.2 | F4:39:09:02:F2:6C (HP) |
HP ProDesk 600 G4 SFF desktop (admin PC) | Confirmed | (none) | mDNS/PTR name administrator-HP-ProDesk-600-G4-SFF-IDS-APJ.local; HP OUI; host firewall closed |
| 192.168.1.5 | 7C:B6:8D:20:0E:32 |
Unidentified endpoint — possibly an AP32, an IP phone, or another PC | Unknown | (none) | Alive at L2 (ARP), answers ICMP, no listening TCP ports probed |
| 192.168.1.6 | 04:7C:16:A8:06:FA (Realtek) |
This PBX laptop | self | — | Serve.ps1 console host |
Ports probed: 21,22,23,25,53,80,110,135,139,143,443,445,514,548,830,902,993,3389,5060,5901,8080,8443,9100.
4. Per-device detail¶
4.1 SRX — 192.168.1.1 (the edge, and the most feature-rich box for us)¶
- Roles proven live: default gateway; DHCP server (handed this laptop
.6, 24 h lease, obtained 2026-07-18 12:52 → expires 2026-07-19 12:52); NTP server (answered a client query on 123/udp). - ⚠️ NTP reply came back stratum 0 = not synced to an upstream clock. Air-gapped means no internet time source; if the SRX is meant to be the campus clock (plan D4) it likely needs a local-clock (
server 127.127.1.0) or GPS reference so it serves a defined stratum. Worth verifying. - Management: SSH (
OpenSSH_7.5→ older Junos, normal for SRX300/320), J-Web on 443 (TLS 1.2 / AES256), NETCONF on 830. - Serial (from J-Web cert):
CV3225AX0113.
4.2 EX2300-C ×2 — 192.168.1.3, 192.168.1.4¶
- Two identical units (Juniper OUI
E8:A5:5A, consecutive MACs) — matches the 2× EX2300-C inventory. - Modern, CVE-patched Junos (
OpenSSH_9.7 with CVE-2024-6387,CVE-2024-39894 fixes) — good; these are current. - Management: SSH (22), telnet (23) — open, NETCONF (830). No J-Web observed (443 closed) — either web-management off or not licensed on the switch.
- Not yet a Virtual Chassis that we can prove from here — need
show virtual-chassis(§7). Both having their own management IP on the flat subnet is consistent with two independent switches right now (not yet VC'd per plan T0.1).
4.3 HP ProDesk 600 G4 SFF — 192.168.1.2¶
- A Windows desktop (
administrator-HP-ProDesk-600-G4-SFF-IDS-APJ), HP OUI, host firewall closed to all probed ports. Likely the admin/management workstation on the bench. Not part of the emergency data path.
4.4 Unidentified endpoint — 192.168.1.5¶
- Alive (ICMP + ARP), OUI
7C:B6:8D, no listening services. Could be a Mist AP32 (cloud-managed APs expose no local TCP), an IP phone, or another PC. Resolve via the switch (show ethernet-switching table+show lldp neighborswill name it and give its port). Flagged as the top unknown.
5. Services & management-surface map¶
| Capability | Where | Status | Use for UPES-ECS |
|---|---|---|---|
| NETCONF (830/SSH) | SRX, EX#1, EX#2 | Open on all three | The transport for T3.1 network-health panel, T3.2 caller-location, T3.3 evac-mode, T7.3 config backup — ready today, just needs a read-only user + collector scripts. |
| SSH (22) | all three Junos | Open | show/config-as-code, backups. |
| J-Web (443) | SRX only | Open, TLS 1.2 | GUI inspection of the SRX. |
| DHCP | SRX .1 |
Serving (24 h leases) | Where phone/laptop reservations live (Juniper.md §2). |
| NTP (123) | SRX .1 |
Serving, stratum 0 | Intended campus clock (plan D4) — verify it's actually synced/defined. |
| DNS | WARP on the laptop (127.0.2.2) |
n/a | The laptop resolves via Cloudflare WARP, not the SRX — so the SRX's own DNS zone is invisible from here. Unrelated to the Juniper gear. |
| SNMP (161) | — | public refused everywhere |
Good (not wide open). If we want SNMP/RMON telemetry later it must be explicitly enabled with a community/user. |
| Telnet (23) | EX#1, EX#2 | Open | Hardening item — disable in favour of SSH-only (plan T2.7). |
6. What this means (readout against the plan)¶
- We are at Phase 0 (flat pilot). No VLANs, no Virtual Chassis, no LLDP-MED/CoS/PoE policy provable from here. The whole Tier 0 fabric (VC, voice VLAN, zero-touch phones, RTP priority, PoE budgeting) is still to build.
- The differentiators are unblocked at the transport layer. NETCONF being live on all three boxes means T3.1/T3.2/T3.3 are buildable now — the only missing pieces are a read-only Junos user and the collector scripts (
deploy/juniper/collectors/). - The APs are the gap. With the 2× AP32 absent, the Wi-Fi 111 path isn't live on this segment. Commissioning them (plan §7) is prerequisite to the wireless tiers.
7. How to extract more (read-only) — the next data we can't get without login/elevation¶
Everything above was unauthenticated. The deep inventory needs either a read-only login or a local elevation, both still read-only:
A. Sniff one LLDP frame (no switch login; needs local admin on the laptop). The EX sends LLDP every ~30 s on the laptop's port; capturing one reveals the switch hostname, the exact port the laptop is on, and the switch's mgmt address — the single highest-value missing datum. Run in an elevated PowerShell:
pktmon start --capture --pkt-size 0 -c 1
# wait ~35 s for an LLDP frame, then:
pktmon stop
pktmon pcap PktMon.etl -o lldp.pcap # open in Wireshark; filter: lldp
B. Read-only Junos login (needs a read-only user/creds). With a class read-only user (or view-only NETCONF), these give the full picture — all read-only, zero config change:
# Identity / hardware
show version | show chassis hardware | show chassis routing-engine | show system uptime
# The fabric
show virtual-chassis # are .3/.4 a VC or independent?
show vlans | show ethernet-switching table # what MAC is on which port -> identifies .5
show lldp neighbors # what's plugged into each port (phones/APs/PBX)
show interfaces terse | show spanning-tree bridge
# Power (the 124 W budget)
show poe controller | show poe interface # draw vs budget, per-port priority
# Edge / services (SRX)
show configuration | display set # full config as set-commands (for git)
show security zones | show security policies | show security alg status # is SIP ALG off?
show dhcp server binding | show ntp status | show ntp associations
show configuration system syslog # where logs go
C. NETCONF collector (read-only, scriptable — the plan's net_health.py seed). PyEZ/ncclient against 830 with a read-only user pulls get_poe_interface_information, get_lldp_neighbors_information, get_interface_information, get_chassis_inventory, get_virtual_chassis_information as structured data — the basis for the Console /network panel (plan T3.1).
None of A/B/C changes anything on the devices. If/when you want the deep pull, provide a read-only login (or run B yourself and paste the output) and this doc gets a §4 addendum with the real hardware/PoE/neighbor data.
8. Security & hygiene observations (read-only findings)¶
- Telnet open on both EX switches — plaintext management. Disable, SSH-only (
delete system services telnet) per plan T2.7. - Flat L2, no segmentation — every device (incl. the admin PC) shares one broadcast domain with the (future) phones/PBX. Fine for a pilot; the voice/mgmt/guest VLAN split (plan §3) is the growth path.
- SNMP
publicrefused — good. Keep it explicit-only if enabled later. - SRX NTP stratum 0 — verify the campus clock is actually defined/synced (air-gap has no internet time).
- J-Web exposed on the SRX — fine on a trusted bench; lock to the mgmt network before go-live (plan T2.7).
- Laptop DNS is Cloudflare WARP, not the SRX — a laptop-side VPN, unrelated to the Juniper gear, but it did skew the very first scan (phantom
172.16.16.xhosts via the default route while Wi-Fi was down). Noted so it isn't mistaken for campus infrastructure.
9. Open questions to resolve next¶
- Identify
192.168.1.5(AP32? phone? PC?) — needs switch MAC table / LLDP. - Where are the 2× AP32s? Not on this segment — undeployed, unpowered, or on another port/VLAN.
- Are
.3/.4a Virtual Chassis or two independent switches? —show virtual-chassis. - Physical cabling / port map — which device is on which
ge-0/0/x(feeds T3.2 caller-location). - SRX config specifics — SIP ALG state, zones, DHCP reservations, NTP source.
Read-only discovery. No Juniper device was authenticated to or modified. Reproduce with the
techniques in §7 (elevation) or a read-only Junos login. Pairs with the planning trilogy in
Docs/Juniper*.md.