Recurring intermittent Internet outages lasting 1-4 minutes - Charlotte, NC

evelane

Intermittent Spectrum Internet Outages – Technical Summary

Beginning Jul 16, 7:33:17 AM, I started experiencing recurring intermittent Internet outages on my Spectrum 500/20 Mbps service. The outages typically last approximately 1–4 minutes and recover automatically without rebooting the modem or router. This has been occurring repeatedly over multiple weeks and, recently, multiple times per day.

All equipment has been power-cycled a number of times over the last few weeks. The modem has also been restarted through the Spectrum app.

Equipment

• Spectrum modem: EU2251
• Router/Gateway: Ubiquiti UniFi Cloud Gateway Ultra
• WAN connection: Direct Ethernet connection from Spectrum modem to UniFi WAN port
• WAN Ethernet negotiation: 2.5 Gbps Full Duplex
• Spectrum-assigned public IP during testing: 104.139.XXX.XXX (hidden)
• Spectrum next-hop/default gateway: 104.139.64.1
• Service tier: 500 Mbps down / 20 Mbps up
• Typical measured throughput when connection is functioning: approximately 535 Mbps down / 22 Mbps up
• IPv6 is disabled.

Symptoms

During an outage, all Internet connectivity stops for approximately 1–4 minutes and then recovers automatically.

The Ethernet connection between the EU2251 modem and the UniFi gateway does not drop. The WAN port remains linked at 2.5 Gbps Full Duplex.

The EU2251 modem also does not show an obvious loss of DOCSIS connectivity. Both modem LEDs remain solid blue, exactly as they appear during normal operation. I have previously observed the modem's 2nd LED pulse blue/white during known Spectrum outages, but that does not occur during these events.

After connectivity returns, performance immediately returns to normal. A speed test after an outage measured approximately 535/22 Mbps.

UniFi WAN Monitoring

The UniFi Cloud Gateway Ultra independently detects these events and records:

• Packet Loss Detected
• Internet Down
• Internet Restored
• Temporary Internet Disconnection
• Multiple Internet Disconnections

The gateway uses multiple independent Internet verification mechanisms:

1. ICMP/Ping to ping.ui.com
2. DNS verification against 1.1.1.1
3. DNS verification against 8.8.8.8

During normal operation all three report 100% availability with approximately 23–28 ms latency.

The UniFi gateway recently reported only 99.17% Internet uptime over a 24-hour period, consistent with the multiple short outages being experienced.

Direct Testing During the Most Recent Outage

To isolate the problem, I opened four simultaneous SSH sessions directly on the UniFi Cloud Gateway Ultra and continuously pinged:

1. The router's own Spectrum-assigned public IP: 104.139.XXX.XXX
2. Spectrum's next-hop/default gateway: 104.139.64.1
3. Google DNS: 8.8.8.8
4. Cloudflare DNS: 1.1.1.1

During the outage:

• 104.139.XXX.XXX continued responding.
• 104.139.64.1 stopped responding.
• 8.8.8.8 stopped responding.
• 1.1.1.1 stopped responding.

This testing was performed directly from the WAN router itself rather than from a LAN or Wi-Fi client.

The WAN Ethernet link remained up throughout the event.

evelane

Captured Packet-Loss Event

The continuous ping to Spectrum's next-hop gateway produced the following immediately before and after one outage:

64 bytes from 104.139.64.1: icmp_seq=9760 ttl=64 time=16.0 ms
64 bytes from 104.139.64.1: icmp_seq=9761 ttl=64 time=11.8 ms
64 bytes from 104.139.64.1: icmp_seq=9762 ttl=64 time=8.59 ms



64 bytes from 104.139.64.1: icmp_seq=9870 ttl=64 time=32.7 ms
64 bytes from 104.139.64.1: icmp_seq=9888 ttl=64 time=26.5 ms
64 bytes from 104.139.64.1: icmp_seq=9889 ttl=64 time=23.9 ms

The sequence numbers show approximately 107 consecutive ICMP requests without a response between sequence 9762 and 9870.

A single response was then received at sequence 9870, followed by approximately another 17 unanswered requests before replies resumed consistently at sequence 9888.

At the same time, continuous pings to both 8.8.8.8 and 1.1.1.1 also stopped receiving responses.

This indicates that connectivity was being lost to Spectrum's first Layer-3 next hop, not merely to destinations farther out on the Internet.

Examples of Recorded Outages

UniFi has recorded numerous events, including:

• Jul 16 7:33 AM – Internet Down; restored 7:36 AM
• Jul 17 8:17 PM – Internet Down
• Jul 19 2:47 PM – Internet Down; restored 2:49 PM
• Jul 23 5:01 AM – Internet Down; restored 5:04 AM
• Jul 24 2:15 AM – Internet Down; restored 2:16 AM
• Jul 24 6:15 PM – Internet Down; restored 6:18 PM
• Jul 24 9:46 PM – Internet Down; restored 9:47 PM
• Jul 24 10:42 PM – Internet Down
• Jul 25 1:31 AM, 1:53 AM, 4:24 PM and 4:37 PM
• Jul 26 1:33 AM, 1:45 AM, 2:08 AM, 1:21 PM and 3:30 PM
• Jul 27 5:21 AM, 11:34 AM, 11:50 AM, 11:53 AM, 12:00 PM and 5:29 PM
• Jul 28 7:18 PM
• Jul 29 11:21 PM
• Multiple additional outages have occurred since then.

The events occur at varying times of day and do not appear consistent with a normal DHCP lease renewal or scheduled event.

Important Observations

The following have been confirmed:

• This is not a DNS-specific issue. During an outage, direct ICMP connectivity to public IP addresses fails.
• This is not a Wi-Fi issue. Testing is being performed directly from the wired WAN gateway.
• The UniFi router remains operational.
• The UniFi router retains its Spectrum-assigned WAN IP.
• The Ethernet link between the router and EU2251 modem remains up at 2.5 Gbps.
• The EU2251 LEDs remain solid blue.
• Spectrum's first-hop gateway becomes unreachable.
• Multiple unrelated external IP addresses simultaneously become unreachable.
• No equipment needs to be rebooted for service to recover.
• Normal latency and full throughput return immediately following recovery.

Requested Spectrum Investigation

Please review the EU2251 modem and CMTS telemetry/history for the times of these events rather than relying only on its current status, since the connection appears completely normal between outages.

In particular, please check:

• DOCSIS upstream and downstream RF signal history
• Upstream SNR
• Downstream SNR
• T3 timeout history
• T4 timeout history
• Partial Service events
• Lost MDD events
• Correctable and uncorrectable codeword/FEC errors
• Upstream ranging events
• CMTS modem flap history
• CMTS event history for this modem
• Noise/ingress affecting the upstream channels
• Spectrum tap/drop/connector condition
• Modem firmware or forwarding issues with the EU2251
• Any CMTS or node-side events corresponding with the outage timestamps

Because the Ethernet link remains established while the Spectrum next-hop gateway becomes unreachable, I would specifically like the modem's historical DOCSIS telemetry and CMTS-side records examined for the outage timestamps.

If the RF/CMTS history is clean, I would also like the EU2251 modem replaced to eliminate an intermittent modem forwarding or hardware/firmware problem.

The problem is intermittent, so checking the connection only while it is functioning will likely show normal signal levels and throughput. The historical modem/CMTS data corresponding to the documented outage times is important to diagnosing the issue.

evelane

GatewayEventsJuly.txt

Event log attached.

William_M

Hey @evelane, welcome to our community! I see signal issues from 8/01 which caused some timeouts and offline events for ~3 minutes on two occasions, and 6 minutes offline on 7/27 (11:53-11:59 AM), but otherwise it has been online with healthy signal strength. Would you be able to connect just a single device straight to the modem with ethernet (bypassing your router,) reboot the modem to restore internet, and test if you continue to have problems that way? You are certainly welcome to exchange your modem at a Spectrum store if you want to try that first.

evelane

Thank-you so much for taking a look, and for the reply, @William_M.

I've ordered a CM3000 which I'll get provisioned today. I just wanted more visibility to the other side.

I ran all my tests directly on the Unifi CGU (via SSH sessions) which is a device connected directly to the modem via Ethernet. But, I understand.

I'll prepare a machine to record data, as well as regularly scrape the CM3000 screen (unfortunately it doesn't support SNMP).

There were eight hours between outages yesterday (Aug 1, 11:48:31 AM and Aug 1, 7:54:38 PM, which is actually a notable improvement) so I'll need to plan an outage window to run the bypass test, as taking down the internet for an extended period breaks a lot of things.

Thanks again. I'll give you an update after the test.

Best regards,
Chris

evelane

Baseline Readings After Installing Replacement CM3000

I replaced the Spectrum-provided EU2251 with a new NETGEAR CM3000. No other changes were made to the network, coax, Ethernet connection, or router. These readings were captured shortly after activation to establish a baseline before another outage occurs.

Downstream DOCSIS 3.0 (32 channels):

• All 32 channels locked
• Power levels approximately +1.9 to +3.8 dBmV
• SNR approximately 42.5–44.2 dB
• Correctable codewords: 0
• Uncorrectable codewords: 0

These downstream levels appear very clean and consistent across all channels.

Downstream DOCSIS 3.1 OFDM:

• OFDM channel locked
• Frequency: approximately 617 MHz
• Power: approximately +0.68 dBmV
• MER/SNR: approximately 42.5 dB
• Uncorrectable codewords: 0

Upstream:

• Two ATDMA/SC-QAM upstream channels locked
• ATDMA transmit power: approximately 45.5 dBmV
• DOCSIS 3.1 OFDMA upstream channel locked
• OFDMA transmit power: approximately 43.8 dBmV

The modem's own diagnostic page currently rates the overall cable connection, downstream power/SNR, upstream connection, and upstream power as Good.

Event log:

There were several synchronization/ranging events during the initial modem activation and provisioning, which is to be expected and I am disregarding as startup-related. However, after the modem was operational, it recorded a T3 timeout:

Sun Aug 02 11:26:54 2026

Started Unicast Maintenance Ranging - No Response received - T3 time-out

This occurred after normal operation had been established, so I am documenting it as part of the baseline.

At this point, RF levels appear very good. I am leaving the configuration unchanged and will monitor for another Internet outage. If another outage occurs, I will compare its exact timestamp against the CM3000's DOCSIS signal levels and event log, particularly looking for T3/T4 timeouts, ranging failures, loss of OFDMA/upstream channels, or other DOCSIS events.

The CM3000 shipped with FW v6.01.04. After the Spectrum initialization I noted that it was downgraded to v6.01.03. This was simply an observation.

I'm writing a scraper to continuously collect and log the following from the CM3000:

Metric	Baseline
Downstream power, all channels	+1.9–+3.8 dBmV
Downstream SNR/MER	42.5–44.2 dB
OFDM power	+0.68 dBmV
OFDM MER	42.5 dB
Correctables	0
Uncorrectables	0
Upstream power	45.5 dBmV
OFDMA power	43.8 dBmV
Upstream channel lock	Locked
OFDMA lock	Locked
Modem uptime	—
T3 timeouts	One post-activation
T4 timeouts	None seen
Partial-service/ranging events	None established

Best regards,
Chris

evelane

@William_M we've had no outages since the CM3000 modem was installed. I'll be taking the EU2251 modem to a Spectrum office for replacement.

Best regards,
Chris

William_M

Sorry for the delay, but great to see it's working better now! Please let us know if you have any further problems with the new modem.