※日本国内案件については、元請けITベンダー様経由での技術支援を基本としています。
Proven Troubleshooting and Recovery Cases in Enterprise Networks

Proven Network Troubleshooting Cases

These network troubleshooting cases were resolved in enterprise production environments.

Cisco SD-WAN

We resolved operational limitations in Cisco SD-WAN environments by introducing automation.

Tasks difficult to perform through the GUI were implemented using Python.
Legacy TeraMacro procedures were translated into Python using generative AI, including automated configuration backup operations.

To prevent operational mistakes, a safety mechanism was implemented:
if the expected management IP address does not exist in the configuration, the script automatically stops.

All intellectual property must remain with the client.
Therefore, we use client-owned generative-AI environments when generating scripts.

Any modern tool must be usable by anyone.
If only specialists can operate it, it has limited value.
TeraMacro training costs are extremely low.People simply buy used routers on Yahoo Auctions (typically only a few thousand yen).

During pre-deployment validation, we discovered that the legacy BGP command “allow-as in” cannot be implemented in SD-WAN.
We resolved this using redistribution and route filtering.

IOS-XE 9200/9300 Switching

We resolved an issue where the command
“no spanning-tree vlan xx”
could not be applied.

The issue was solved using BPDU filter and BPDU guard.
Other members had been unable to resolve it before our intervention.

AWS

When web filtering was enabled, uploads failed with a probability of roughly 19 out of 20 attempts.

Root cause:

  • Global IP changed mid-session due to virtual server relocation
  • Non-DNS-based algorithm
  • Packet fragmentation preventing Layer-7 inspection

This was confirmed through packet capture analysis.

VPN / IPsec

We identified incorrect hardware selection in a failed data-leak-prevention deployment.

We resolved a billing-related issue in an on-demand VPN circuit where packets continued arriving after communication completion due to IPsec confirmation behavior.

We also resolved:

  • QoS not functioning with IPsec
  • MTU issues caused by key-length changes
  • Customer concerns about unencrypted voice packets (disproved through waveform analysis)

Layer 7-2 (Transparent IPS/WAF)

In transparent IPS/WAF environments, HSRP hello frames and R-STP BPDU frames did not pass by default on certain platforms.

Impact:

  • HSRP Active-Active state
    I “came up with” the idea on the spot to roll back using an RJ-45 J-J connector.
  • Up to 5 minutes of network outage
    This issue had remained unresolved for three years.

We also discovered in advance that Auto-MDI becomes disabled when a transparent IPS loses power, which can cause link failure with fixed-speed devices.

Layer 4-3 (Load Balancers / Firewalls)

We discovered source-port exhaustion and TIME_WAIT reuse issues when SNAT was enabled on load balancers.

We also resolved:

  • RedHat memory exhaustion caused by RST-terminated health checks
  • Embryonic timeout issues
  • TraceRoute being SNATed by default
  • Firewall uRPF alerts triggered by TraceRoute

All confirmed via packet capture.

Layer 3 Routing

We discovered incorrect QoS + NAT implementation described in Cisco documentation.

ACLs referencing IP addresses did not produce expected results.
Using port-based ACLs resolved the issue.

Wireshark graphs changed from a sawtooth pattern to a straight line, proving QoS effectiveness.

We also:

  • Identified QoS misconfiguration with priority queue
  • Predicted CPU overload during NAT migration
  • Discovered stateful NAT left unconfigured for five years
  • Confirmed PIM multicast and HSRP interoperability

Layer 2 Switching

We proved that some switches configured for untagged VLANs forward all tagged frames regardless of VLAN ID.

We also:

  • Prevented STP root-bridge takeover during switch addition
  • Resolved multicast MAC conflicts between IGMP and BPDU

Layer 1 / Wi-Fi

We resolved Wi-Fi multicast performance degradation caused by lack of Layer-1 ACK.

On Cisco WLC, converting multicast to unicast resolved the issue.

We also:
Resolved MAC-flap instability
Addressed TEMPEST concerns by proving no electromagnetic leakage