GhostMesh
What data is in a GhostMesh diagnostic report?
Investigating a stalled connection needs timing, the failure stage and technical counters. However, a diagnostic report can mean several different files. This article explains the compact automatic SRSP incident format and why its properties do not apply to every log or packet capture.

At a glance
- An automatic incident report contains a limited set of technical fields, including versions, identifiers, event timings and counters. It is not a packet payload recording.
- An exported log, screenshot or packet capture has a different scope and needs its own review before sharing.
- The report format does not describe every data-processing activity of the service. The receiving API handles a request; consult the privacy policy for processing terms.
Establish which report you have
GhostMesh includes diagnostics for connection setup stages and a mechanism for capturing incidents during an active session. Setup diagnostics help locate a failed connection step. Incident recording collects a bounded sequence of facts around stalls or repeated reconnects. This article describes the latter format based on the current Windows and Android client implementations and receiving API.
A text application log, manually exported archive, error screenshot or packet capture does not automatically have the same properties. A diagnostic filename does not make a file suitable for public sharing. Its contents depend on how it was obtained and which application version produced it.
| Material | What to consider |
|---|---|
| Compact automatic SRSP incident | Defined fields and format validation before sending |
| Manually exported log or archive | May include extra lines, paths and context; inspect it separately |
| Screenshot or packet capture | May expose personal details or network addresses; review its scope separately |
Fields in a compact incident
The client rebuilds an incident from allowed fields instead of forwarding an arbitrary original object. It adds the platform, application version and selected server identifier when available. Network type is a category such as Wi-Fi, cellular or unknown. The current Windows adapter can leave this field as unknown.
An incident contains a random identifier, event type, core build, timestamps and a bounded measurement sequence. This can include transmitted and received byte counts, an RTT estimate, reconnect attempts and missing heartbeat replies. Supported versions also include aggregate TCP, UDP and DNS counters. A field's presence in the schema does not mean every version always supplies it.
| Group | Purpose |
|---|---|
| Versions, platform and server | Relate the failure to a particular client and connection setup |
| Identifiers and event timings | Associate relevant stages and avoid duplicate incidents |
| Bytes, RTT and reconnects | Distinguish missing replies, slowdowns and repeated attempts |
| TCP, UDP and DNS aggregates | Identify which class of operation stopped receiving expected replies |
What this format does not contain
The allowed automatic incident schema does not provide fields for message contents, web page bodies, passwords, connection keys, a complete browsing history or the destination addresses of individual user connections. DNS counters describe operation counts and outcomes, rather than a list of queried domains. Byte counts describe data movement, not the content transferred.
That statement is specific to the incident format. It does not mean the receiving server cannot see the source address of an HTTP request, or that other logs have the same scope. Nor should de-identification be interpreted as a guarantee that no technical information can ever be correlated.
How a report reaches the server
In the implemented mechanism, completed incidents can enter a bounded local queue. If the API is unavailable, a temporary delivery failure can leave the entry for a later attempt. Receipt therefore need not coincide with the original failure. Sending does not require the user to manually attach every incident to a support conversation.
The receiving API validates allowed fields and limits request size and frequency. Submission uses an authenticated request. The server can use the request's source address to determine a country and network autonomous system; incident tables include these network categories. The absence of a full IP address in the client report body does not mean the service never processes an IP address.
A local queue is different from server storage
Local queue limits prevent diagnostic files from growing indefinitely when connectivity is unavailable. The API separately defines deletion periods for detailed records and retention of aggregates. These are different layers: deleting a local file is not a request to delete an already delivered server record.
Do not assume one retention period covers every diagnostic material. An archive attached to a support request and an automatic incident can have different handling rules. Check the privacy policy and ask support about the specific file type for storage and personal-data handling terms.
Before manually sharing a file
Start with a description, the time and time zone, and the application version. If support needs a file, clarify its type and how to obtain it. Do not send the entire application folder or a full traffic capture by default: extra material adds unnecessary information and complicates investigation.
- Check the filename and how it was produced: an automatic incident, an application log or a packet capture.
- Inspect text and images for tokens, keys, email addresses, usernames, local paths and personal messages.
- Remove unrelated applications and notifications from screenshots. Ask support first if unsure which report fields can be redacted.
- Use an official support channel rather than posting the original file in a public conversation.
- Include a short reproduction sequence and repeat-check result to give the report useful context.
What an incident does not establish
Missing incoming bytes or repeated reconnects help locate a failure, but do not automatically identify who caused it. The device, Wi-Fi, carrier, route or remote service can all be involved. A conclusion needs symptoms and multiple observations considered together.
Collect a minimal description first, then use a connection check appropriate to the symptom. If the browser works but a call does not, record that separately. This is more useful than saying VPN does not work and lets the investigation proceed without sending the contents of personal conversations.
Sources
Start with a connection check
Choose a check matching the symptom and record the result before sharing diagnostic files.
Open diagnostics