Blog Post

Windows IT Pro Blog
6 MIN READ

Collecting Delivery Optimization logs the easy way

Andy_Rivas's avatar
Andy_Rivas
Icon for Microsoft rankMicrosoft
Aug 18, 2026

You've deployed Delivery Optimization settings and perhaps even a Microsoft Connected Cache node to reduce network traffic on Patch Tuesday. But things aren't working as expected. You aren't seeing good content peering efficiency. You notice peers reaching across the internet to other sites or find that content isn't coming from a Connected Cache. What do you do next?

If you've ever had to troubleshoot a content delivery issue with Delivery Optimization or Connected Cache and or needed to send logs to Microsoft support — or just wanted to dig in yourself — let's talk about a free tool that can make the process easy.

The Delivery Optimization Troubleshooter is a Microsoft diagnostic tool designed to quickly help you identify why Delivery Optimization or Microsoft Connected Cache isn't working as expected. It collects configuration details, network diagnostics, service status, performance data, and logs into one support-ready bundle, making troubleshooting faster and easier.

How to use the Delivery Optimization Troubleshooter to collect diagnostic and debugging information

The first step is to download and install the Delivery Optimization Troubleshooter from PowerShell Gallery on the device you want to debug and run the following command using the PowerShell console:

Install-Script -Name DeliveryOptimizationTroubleshooter

To collect the logs, all you have to do is run this one-line command:

DeliveryOptimizationTroubleshooter.ps1 -GenerateSupportBundle

This command collects a full snapshot: Delivery Optimization service logs, configuration, network info, and event logs — packaged into a zip file and ready to share. Pretty easy, right?

What if you want to reproduce a live issue? Add ReproduceIssue to the GenerateSupportBundle command as shown in this example:

DeliveryOptimizationTroubleshooter.ps1 -GenerateSupportBundle -ReproduceIssue

How the Delivery Optimization Troubleshooter works

At a high level, here is what this troubleshooter does and how it can be used to debug a live issue:

  1. The tool switches Delivery Optimization tracing to verbose logging and increases log limits.
  2. The tool stops the Delivery Optimization service, wipes old log files (creating a clean slate), and restarts with verbose logging enabled.
  3. You reproduce whatever was failing (e.g. trigger the download or run Windows Update), then press Enter.
  4. The script stops Delivery Optimization again, captures the verbose reproduction logs, disables verbose logging, and moves on.

Here's a detailed description of what happens throughout the support bundle generation process:

Step

What it captures

1

Existing Delivery Optimization logs

2

Delivery Optimization service status

3

Delivery Optimization performance snapshot

4

Delivery Optimization configuration details

5

Network configuration details (ipconfig /all)

6

Network List Manager (NLM) data (which networks Windows sees and their states)

7

NLM cost data (shows whether the connection is metered)

8

Proxy settings

9

DoSvc registry key

10

Troubleshooter diagnostics output

11

Interactive pause (waits for you to reproduce the issue)


In the final step, the tool creates the zip file and stores it in %TEMP% as dosvc-diag-[username]-[machinename]-[timestamp].zip.

Here are the typical files produced after running the troubleshooter, and what each one tells you:

File

What's in it

troubleshooter-output.txt

Structured pass/fail diagnostic checks (the fastest file to start with)

status-dosvc.txt

Per-job download breakdown: every file Delivery Optimization has touched, bytes from each source (HTTP, peers, Microsoft Connected Cache), download duration, caller app

logs-dosvc-repro.txt

Verbose logs captured during your reproduction window

logs-dosvc-existing.txt

Pre-existing Delivery Optimization logs (large, but useful for history before the issue)

network-nlm-data.txt

Full NLM network data (connected networks, interface states)

network-ipconfig.txt

Full ipconfig /all output

config-dosvc.txt

All Delivery Optimization policy values and their source (MDM, GPO, or default)

dosvc-registry.reg

Raw DoSvc registry key export

network-nlm-cost.txt

Connection cost (unrestricted vs. metered; Delivery Optimization throttles aggressively on metered)

network-proxy-settings.txt

WinHTTP + IE proxy config

perfsnap-dosvc.txt

Monthly bandwidth totals by source: HTTP, LAN peers, Microsoft Connected Cache, internet peers

I have the output from the Delivery Optimization Troubleshooter. What's next?

Start with the troubleshooter-output.txt file. It runs structured checks and surfaces failures immediately. Example from a real run:

Name : Download Mode
Result : Pass
Details : LAN - 1

Name : CacheHost server configured
Result : 10.1.1.101

Name : Download from CacheHost server
Result : Fail
Details : Unable to connect to the remote server 10.1.1.101:80

 

In this example, the last detail instantly tells you Microsoft Connected Cache is configured, but is not reachable on the device.

The perfsnap-dosvc.txt shows your monthly Delivery Optimization bandwidth split at a glance:

DownloadHttpBytes : 27,171,761,913 ← 27 GB straight from CDN
DownloadCacheHostBytes : 0 ← 0 bytes from MCC this month
DownloadLanBytes : 747,999,564 ← ~700 MB from LAN peers

If DownloadCacheHostBytes is 0 on a machine that should download from a Microsoft Connected Cache, this confirms what the Troubleshooter output had shown.

Status-dosvc.txt tells you per-download where bytes came from, which caller app triggered it (PredefinedCallerApplication: IntuneAppDownload, WindowsUpdate, etc.), and how long the download took. This can be useful for showing that a specific content type isn't routing through Microsoft Connected Cache.

When analyzing the bundle, here are the things I look for first:

  • HRESULTerror codes: These tell you what failed (e.g., 0x80D03002 = no peers found, 0x80D02002 = timeout reaching CDN). See Delivery Optimization Return Values for more details.
  • DownloadMode: This value confirms whether P2P/Microsoft Connected Cache is enabled. See the Download Mode documentation for more details.
  • GroupID: Misconfigured group IDs are a surprisingly common cause of peers not finding each other. For more information, see the Group ID documentation.

Note: The current DeliveryOptimizationTroubleshooter version 1.3.0 contains a built-in list of 72 error definitions. There is a JSON array embedded as $errorsObj inside the script's Get-DOErrorsTable function. With a standard installation, you can inspect it at: C:\Program Files\WindowsPowerShell\Scripts\DeliveryOptimizationTroubleshooter.ps1 The source and latest version are available at aka.ms/do-fix

Below are some real-world examples from customer logs that demonstrate cases where content came from Connected Cache or didn't come from Connected Cache.

Example 1: 100% downloaded from Connected Cache

caller: IntuneAppDownload
cacheHost: GP:192.168.10.5
cdnIp: [192.168.10.5]:80; X-CCC:...; X-CID:10000
File: 157504
DOINC: 157504
conns: [CDN: 0, DOINC: 1]
timeMs: 1482
sessionTimeMs: 619

All 157,504 bytes came through Microsoft Connected Cache, confirmed by the private "cdnIp" (Delivery Optimization sees Microsoft Connected Cache is just another CDN) and X-CID:10000 (X-CID 10000 = Enterprise Connected Cache). This is a very clear example of the device downloading 100% of the content from a local Connected Cache server.

Example 2: No downloads came from Connected Cache

caller: MDMSW Job
cdnUrl: https://naprodimedatapri.azureedge.net/IntuneWindowsAgent.msi
cdnIp: [13.107.246.73]:443
cacheHost: GP:192.168.10.5
File: 7901184
DOINC: 0
conns: [CDN: 1, DOINC: 0]
timeMs: 94571
sessionTimeMs: 1176

Intune-related content bypassed using Connected Cache for downloads, but why did the client not use Connected Cache even though it was configured? Hint: What port was used for the download? 🤔

Digging deeper into the logs you will see more details on the actual download event, which confirms the suspicion.

DownloadError: hr = 80072efd, httpCode = 0,
url = https://192.168.10.5/...IntuneWindowsAgent.msi
…
Unresponsive cache host 192.168.10.5
ErrorCode: -2147012867 # 0x80072EFD
Ban count for 192.168.10.5 is 1
…
Establishing HTTP Connection to
https://naprodimedatapri.azureedge.net/IntuneWindowsAgent.msi
…
cacheHost = 192.168.10.5
bytes: [File: 4995, CDN: 4995, DOINC: 0]
…
Connecting to ... naprodimedatapri.azureedge.net, IP: []:443
HTTP download request complete(hr = 0)

Further evidence can be seen in the logs of the download source:

{CHttpPeerConn::Initiate}
initiating HTTP connection, IP: 13.107.246.73

{CConnMan::ConnectionComplete}
connected to https://naprodimedatapri.azureedge.net/IntuneWindowsAgent.msi
(httpConnection)

{CHttpPeerConn::_OnComplete}
HTTP connection: [13.107.246.73]:443

Aha! The download happened over https, but why did it not come from the Connected Cache? The key is the download error (DownloadError: hr = 80072efd) which is ERROR_WINHTTP_CANNOT_CONNECT. The bottom line is the Connected Cache server is not configured for https downloads, a requirement for Intune content. For more details, see Reminder: Intune-managed Win32 app delivery will be HTTPS-only, affecting Connected Cache customers.

There are many more examples I could share, but I hope this has been helpful to get you started. What Delivery Optimization issues are you seeing most in your environment? Drop a comment — we would be happy to help!

 


Continue the conversation. Find best practices. Join the Windows Tech Community.

For updates on new releases, tools, and resources, bookmark the Windows IT Pro Blog and follow @WindowsUpdate on X. For the latest on how Windows is evolving for organizations and the people who support them, follow Windows at Work on LinkedIn.

Updated Oct 20, 2026
Version 2.0

5 Comments

  • My latest issue is that I wanted to monitor GroupID based "consumption" via WUfB Reports, but the obfuscated GroupID's shown in the reports mostly remain a mystery. Can you please shed some light on what's going on?

     

    I've got a defined GroupID in every subnet and policy telling the endpoints to use it. Out of the 9 displayed obfuscated GroupID's, only one of them pops out of the hashing process to reveal itself as our AD Domain SID. None of the others match DHCP, Entra tenant, AD site names, or DNS domains. Unfortunately, the AD Domain SID one isn't the largest.

    I decided to try and find out what the largest obfuscated GroupID was. It has around 1,600 endpoints in it. I tracked down several devices with this obfuscated GroupID and found their locations to be random from my perspective. Each DHCP subnet confirmed as having a GroupID in the scope options.

    I then focused on one device and grabbed its DO log. There I found a completely different obfuscated GroupID to the one associated in the WUfB report. Low and behold, that obfuscated GroupID in its log was an exact match for the DHCP scope option-supplied GroupID.

    Looking at the WUfB report data, an agent captured that the sample device had two different obfuscated GroupID's associated with it:

    Agent finds multiple GroupID's associated to the same device

    In this example the device had 8 rows matching the expected obfuscated GroupID, but just one with the mysterious obfuscated GroupID that 1,600 devices have also been linked to.

     

    Where are these obfuscated GroupID's coming from because they do not match the published documentation for what they should be?

    Clearly there's a pattern, given 1,600 all got the same one. Is it the shared Public IP? That's the only correlation I could guess here after all the other possible matches failed.

     

    Why is this happening? The clients themselves appear to show that they are correctly using the DHCP scope option.

     

    Could this also explain why peer selection was not sticking to GroupID as expected? I've had to lock it down to the subnet restriction, and this prevents clients on the same site/different subnet from peering.

     

    Nearly all endpoints are Windows 11 25H2, kept up-to-date by Autopatch.

    • Andy_Rivas's avatar
      Andy_Rivas
      Icon for Microsoft rankMicrosoft

      @AndrewMcN_SFRS, yeah, you've hit a reporting quirk. The GoupID GUID you will see in the DO log will not be the actual GroupID. There's some obfuscation happening.

      Other than this quirk, do you see peering results as you would expect per group?

      • AndrewMcN_SFRS's avatar
        AndrewMcN_SFRS
        Brass Contributor

        Hi Andy,

        Thanks for replying. Sorry if I wasn't clear. Unless you're talking about some additional obfuscation, I know all about the obfuscation. In fact, I wasted months thinking my decoding wasn't working because I just couldn't find matches to my DHCP GroupID's. That was the story I TL;DR'd above.

         

        I'm saying there's obfuscated GroupID's in there that don't match any of the expected sources. I'm saying there's at least two obfuscated GroupID's being associated with single devices. Shouldn't that be impossible??? WUfB Reports is picking the wrong/unknown/mysterious obfuscated GroupID's. The ones we can't decode. The report tells me nothing I can use.

         

        When I refer to the GroupID in the DO client log, I mean the already obfuscated one. I can see that the DO client appears correct. Yet this other obfuscated GroupID has been attached to the device within the WUfB reports data. The report template shows most of our results with these mysterious obfuscated GroupID's that match none of the standard origins.

         

        In this snapshot, only one of the obfuscated GroupID's actually matches something:

        It's the one that starts "24g1". That's our AD Domain. That shouldn't even be appearing at all. I have unique DHCP GroupID's for every one of our sites. Group Policy tells the clients to use Download Mode 2 (GroupID). I don't even know what that obfuscated GroupID is with the most clients in it. Nothing I hash matches it.

  • SasStu's avatar
    SasStu
    Brass Contributor

    Hey Andy and team, awesome work, thanks! Any chance we could get this into the Intune Diagnostics package?