Skip to content
INSIGHTS & GUIDES · NETCLOUD24

Collecting useful evidence when a Windows application crashes

Collect a targeted user-mode crash dump with controlled storage and permissions. Learn what LocalDumps can capture, how to preserve build context and why sensitive memory files need careful handling.

Read the article ↓
Editorial illustration: How to Capture a Process Crash Dump with Windows Error Reporting

A crash investigation benefits from a dump tied to a precise failure time and application build. An unlabelled memory file collected days later can be difficult to interpret, while an overly broad capture arrangement can fill a server volume or expose sensitive information. Plan the evidence and its handling before enabling collection.

Windows Error Reporting offers LocalDumps for eligible user-mode crashes. The method below targets one illustrative executable, LedgerWorker.exe. It is intended for the authorised support operator; Netcloud24 customers should provide the symptoms and timing, while the managed team arranges appropriate server-side diagnostics.

Decide whether this is the right capture method

Confirm the executable path, version, architecture and account. Distinguish an unhandled crash from a service that never starts, an application that hangs, a normal error exit and a process killed by another component. LocalDumps does not automatically capture every type of failure.

Ask the application support owner what evidence is needed. A minidump may contain enough information for an exception analysis; a full dump retains more process memory and can be much larger. Neither is automatically useful without matching binaries, symbols or application expertise.

Applications with their own custom crash reporting and systems configured for automatic debugging can have limitations for this mechanism. Review Microsoft's documented conditions before treating a missing dump as a registry fault. A hang may require a separately approved tool or trigger, such as an appropriate ProcDump arrangement.

Agree confidentiality and capacity controls

Process memory can contain credentials, tokens, customer records and decrypted application data. Restrict access, choose an approved transfer route and define deletion or retention before the file exists. Do not attach a dump to a public issue or a broadly visible ticket.

Choose durable local storage with sufficient monitored capacity. Estimate space from the process footprint, desired dump type and retained count, allowing headroom for normal workload activity. A diagnostic feature should not cause a wider service outage by exhausting the system disk.

Use a dedicated directory with narrow permissions. The crashing process needs the documented ability to write there, and authorised analysts need access to collect the result. Ordinary users should not inherit read access simply because the directory sits under a general shared folder.

Preserve existing settings

Inspect both global LocalDumps configuration and an existing executable-specific key. A global policy may already serve other applications. Save the prior values and types through the approved registry backup or configuration record rather than overwriting an unknown arrangement.

$dumpKey = 'HKLM:\SOFTWARE\Microsoft\Windows\Windows Error Reporting\LocalDumps\LedgerWorker.exe'
Get-ItemProperty $dumpKey -ErrorAction SilentlyContinue
New-Item -ItemType Directory -Path 'E:\Diagnostics\LedgerDumps' -Force

Apply the intended directory ACL before enabling collection. Use the actual process identity rather than assuming it matches the administrator performing the setup. Services can use profile locations different from an interactive desktop, which is one reason to choose a deliberate explicit folder.

Configure a bounded per-process request

New-Item -Path $dumpKey -Force
New-ItemProperty -Path $dumpKey -Name DumpFolder `
 -PropertyType ExpandString -Value 'E:\Diagnostics\LedgerDumps' -Force
New-ItemProperty -Path $dumpKey -Name DumpType `
 -PropertyType DWord -Value 2 -Force
New-ItemProperty -Path $dumpKey -Name DumpCount `
 -PropertyType DWord -Value 4 -Force

This example requests full dumps, represented by DumpType 2. DumpType 1 requests minidumps. Select the agreed evidence type rather than always using the largest capture. DumpCount bounds WER's retained files but does not manage copies later transferred elsewhere.

Use the real executable filename, including the extension. A service display name or application-pool label is not an executable. If several applications share a host executable, the scope can be wider than expected and needs review before collection is enabled.

Read the values back and inspect their registry types. Keep to the documented configuration location; copying additional registry recipes from unrelated machines can introduce architecture or scope mistakes.

Validate configuration without creating an outage

Get-ItemProperty $dumpKey | Select-Object DumpFolder,DumpType,DumpCount
Get-Acl 'E:\Diagnostics\LedgerDumps' | Format-List

Confirm the folder exists, appropriate identities can write and read as intended, and unrelated users cannot browse the evidence. Check free-space monitoring and the intended collection period.

A controlled reproduction in a suitable test environment can validate the complete mechanism. Do not deliberately crash production software solely to demonstrate that a key exists. Where production reproduction is necessary for support, it requires the application's explicit incident plan.

Capture the failure with its surrounding evidence

When the relevant crash occurs, record the time, instance, user-visible symptom and application build. Allow writing to complete; large memory files may produce noticeable disk activity. Correlate the file timestamp with application and Windows events rather than selecting whichever dump happens to be newest.

Retain the deployment identifier, relevant logs and available matching binaries or symbols through the approved case process. A filename alone does not establish which build produced it. Check the application's restart or recovery behaviour as part of the incident response.

Transfer the file securely to the authorised analyst. A checksum can confirm transfer integrity, but it does not show the file is free of sensitive data. Keep all working copies under the same access and retention rules.

Investigate missing evidence by eligibility first

If no file appears, confirm an eligible user-mode crash actually happened. An IIS recycle is not necessarily a crash, and a Service Control Manager event describing a stopped process does not establish an unhandled exception.

Then inspect the exact executable name, registry location, value types, folder permissions and free space. Review custom crash handlers and automatic-debugger settings. Check security-product logs through their supported process instead of disabling protection across the server.

If a shared host is involved, identify the particular process instance and workload. Support may need a more precise capture strategy to avoid collecting unrelated customer or application memory.

Close the capture arrangement after investigation

Remove only the temporary application-specific configuration or restore its recorded prior values. Do not delete global settings serving other processes. Registry rollback leaves existing dump files intact, so handle them separately under the retention decision.

Collection is evidence gathering, not a diagnosis. The faulting module named in a simple event is a lead rather than automatic proof of root cause. Microsoft's LocalDumps reference defines the supported mechanism. Related investigations may use IIS worker lifecycle evidence or request tracing.


Was this answer helpful?

« Back
Contact

Write to us

Do you have questions about our services or need a quote? Write to us and we will respond as soon as possible.

Contact form

Fields marked * are required.