Start an order
Root cause analysis / Investigation

copilot.com

Reference RCA-9B998344B43DCB7B

Microsoft Chat website

Measurement point: Stealth NYC, 32 Avenue of the Americas.

Published investigation. Updates run automatically.

01 / Network path

Network path.

Endpoint measurements / ICMP IPv4

Latency over time.

Recorded endpoint latency history is not available for this event. Missing data remains unknown.

Network path / recorded observations

Network path observations.

0 observations

Choose a measurement type to see the closest of the available sampled observations to the selected chart point. Each probe is separate and may follow a different path.

No path observations are available.

The endpoint measurements and collection limitations remain below. Missing data cannot establish a path or a fault.

Same method / same recorded address

Compare the traces.

Compare one selected trace from before the event with one during it. Values are individual replies, not a statistical comparison.

No same-address, same-method pair is available among the selected replay samples. This does not mean no other matching traces were recorded.

RTTs include the return path. Different responders, missing replies and independent probe flows limit what a change can localize. MTR snapshots remain separate in the replay.

Event timeline.

Possible SLA event. Times shown are UTC.

Event opened
UTC
Last event observation
UTC
Recovery observed
UTC

Recorded recovery describes this event, not current reachability.

02 / Assessment

Assessment.

Investigation stateRecovered
Path assessment

Elevated latency observed.

Endpoint round-trip time was elevated during part of the event. The available path observations do not identify where the increase occurred.

What we observed

  • A sustained increase in measured round trip time was observed.
  • Stored traces were compared using matching source and destination addresses.
  • The measured destination address changed across the comparison period.

Recording counts describe collected measurements, not application availability or packet loss. Exact packet paths and the return path remain unverified.

Measurements at a glance.

Baseline RTT6.709 msUsual round trip before the event
Last recorded event RTT6.730 msLast successful event recording
Event p95 RTT22.957 ms95th percentile of successful recordings
Change from baseline+0.021 msLast event RTT compared with the baseline
Successful / expected recordings70 / 70Recorded endpoint samples during the event

RTT is round trip time; p95 describes the upper range of successful recordings. Endpoint measurements differ from individual trace replies and are not an uptime percentage.

Record sources and timestamps
Analysis updated
UTC
Public snapshot published
UTC
Public observation window
– UTC

Event and source timestamps are UTC. The latency chart uses its selected timezone. Page refresh times are separate from measurement timestamps.

03 / Method and limits

Method and limits.

How to interpret the sampled path and its limits

Corroborating the path.

Comparisons use matching destinations before and during the event. ICMP traces use echo probes. TCP/443 traces send SYN probes to port 443 only; they do not complete a TCP handshake or perform a TLS or HTTP exchange.

The Stealth gNMI interface verifies the active route-table choice and maps its primary next hop to the Stealth logical interface, configured attachment and, where applicable, the known LAG member set at each snapshot. Stable Stealth replies, together with mapped interface evidence showing no corresponding errors or sustained capacity pressure, reduce the likelihood of a Stealth forwarding issue. An 80% load warning warrants investigation; it does not prove congestion. Lower interval averages can conceal bursts or queue pressure. Optical readings assess the physical link, not congestion.

What remains unseen.

Probe reply handling can differ from packet forwarding. A slow intermediate reply that disappears at later hops does not locate a forwarding problem. A destination that does not respond to a trace is not thereby shown to be unavailable.

Stealth does not use ECMP in this monitored scope. The route snapshot does not identify which physical LAG member carried a particular independent probe, attest line-card forwarding at that packet instant, prove the exact application packet flow or observe the return path. RTT includes the return path, which may differ. No remote reverse traceroute is available. Tunnels and silent routers may hide parts of the path. Address registration alone does not establish the forwarding operator or responsibility.

References: ICMP reply handling / RFC 1812 · Round trip measurements / RFC 2681 · Full methodology