Forum Discussion
ConfigMgr 2603 PXE suddenly stopped working
ConfigMgr 2603 PXE suddenly stopped working
Looking for some ideas on a PXE issue with Configuration Manager 2603. Everything was working until November 10.
Clients PXE boot and successfully contact the PXE-enabled DP, but then hang while contacting the imaging server and eventually time out before loading the boot image.
Environment:
- Physical DP using WDS
- Enhanced HTTP
- PXE client and DP are on the same subnet/VLAN
- DHCP is on another network using an IP helper
So far I've tried:
- Reinstalling WDS
- New task sequence
- New boot image
- New PXE deployments
- Multiple hardware models
- Disabling firewall/AV
- Verifying RemoteInstall permissions
- Checking WDS, NIC, IIS, and certificates
SMSPXE.log shows the PXE request reaching the DP, but the client stalls afterward and I don't see any useful errors in the DP or MP logs.
I also tested TFTP manually: that does fail for some reason but not sure if by design or not.
5 Replies
- davidbro8Copper Contributor
Given that PXE reaches the DP but hangs when downloading the boot image, I'd focus on TFTP communication between the client and the DP, rather than rebuilding boot images or task sequences.
I'd check these first:
1. TFTP ports: Ensure UDP 69 is reachable, along with the negotiated TFTP transfer ports. Don't just test UDP 69; TFTP uses additional UDP ports for data transfer.
2. IP helpers: Verify the DHCP relay forwards PXE traffic correctly to the DP. Since DHCP is remote, check that no recent network changes affected the relay.
3. WDS configuration: Check that WDS is listening correctly and that the RemoteInstall share is accessible.
4. SMSPXE.log: Look for the boot image being selected and the response sent to the client. This helps establish whether the failure occurs before or during image transfer.
5. Check recent changes around November 10: Windows updates, firewall changes, network firmware, security policies or ConfigMgr hotfixes.
One particularly useful test: Capture network traffic on the DP during a PXE attempt. Filter on:
text
```text
tftp || udp.port == 69
```
Look for unanswered TFTP requests or failed data transfers.
My suspicion is a TFTP transfer problem, potentially involving UDP dynamic ports or a network change.
Question: When the client hangs, does SMSPXE.log show that the boot image was successfully selected and offered, or does it stop before that point?
- ThomasJensenCopper Contributor
I ran a packet trace during my last test and it nothing shows up for tftp || udp.port == 69.
interesting thing is I see a 5 min delay from when the pxe/request comes in and when the client boot reply goes through.
Side note: working with Microsoft we removed WDS and trying to use the deploy pxe without WDS method.
11:59 pm
Packet: Operation: 1 (request), AdrType: 1, AdrLen: 6, HopCount: 0, TransactID: f0180373, BootTime: 28, Addr: mac:00:00:00:00:00:00:00:00:00:00, HostName: , BootFile: , ClientIP: 0.0.0.0, HostIP: 0.0.0.0, ServerIP: 0.0.0.0, RelayIP: 0.0.0.0
Options:
12:04 PM
Client Boot Get ID Info reply: <ClientIDReply><Identification Unknown="0" DuplicateSMBIOS="0" DuplicateMACAddress="0" ItemKey="0" ServerName=""><Machine><ClientID/><NetbiosName/></Machine></Identification><ClientIDInfo ItemKey="0" ClientID="" DuplicateSMBIOS="0" DuplicateMACAddress="0" MatchType="0"/></ClientIDReply>
PXE: mac 0, , SMBIOS ID is NOT a match, MAC Address is NOT a match. SCCMPXE 11/19/2026 12:04:01 PM 3016 (0x0BC8)
PXE: 48:90:F0:94:F7:EC: No valid system records. SCCMPXE 11/19/2026 12:04:01 PM 3016 (0x0BC8)
PXE: 48:90:F0:94:F7:EC: Client machine is UNKNOWN. SCCMPXE 11/19/2026 12:04:01 PM 3016 (0x0BC8)
PXE: mac: Using Management Point: http://mp.domain.com SCCMPXE 11/19/2026 12:04:01 PM 3016 (0x0BC8)
- ThomasJensenCopper Contributor
The current TFTP settings are 1456 block size and 8 window size. I tried lowering both values to see if it made any difference, but the behavior stayed the same. The network team also checked the switch and did not find any dropped packets or other obvious network issues.
At this point, SMSPXE.log isn't showing any errors. I also ran a packet capture against the test device and can see its DHCP traffic, but I'm not seeing PXE or TFTP traffic between the workstation and the DP.
I also tested with PXE-enabled USB media. It boots into WinPE, but then sits at "Searching for machine policy" for about 10 minutes before eventually allowing me to select a task sequence. Once I select the task sequence, it hangs at "Downloading pre-content files" and eventually fails with 0x80040005.
One other odd thing: while it was hung, I tried opening smsts.log, but it was locked because the file was in use. After closing the task sequence/PXE screen and trying again, the log was gone.
I'm not specifically trying to solve the USB boot issue right now, but I wanted to share it because the slow machine-policy lookup and content-download failure could be another symptom of the same underlying communication problem.
- KyleS_85226Copper Contributor
I'm not a PXE expert (we're not using it) but given the timing on 2nd Tuesday release, I would check for NTLM "unsigned" connections to the source of the files. We were bit by this with our legacy proxy infrastructure, which doesn't support signed NTLM, and so we had to push registry value to re-enable it, pending update to the proxy.
HKLM:\SYSTEM\CurrentControlSet\Control\Lsa\MSV1_0 "ClientAllowUnsigndNtlm" ==> 0x1 (DWORD)
Since the client reaches the PXE-enabled distribution point but stalls before WinPE, I’d focus on the boot-file transfer before rebuilding WDS again. TFTP is a normal part of PXE, so the failed manual test is worth investigating, although it doesn’t by itself identify the fault.
Capture traffic during a failing boot on the DP and, if possible, from a mirrored switch port. Check the requested boot filename, server response, retransmissions and any TFTP error; compare those timestamps with SMSPXE and WDS logs. Verify the expected files and Boot.sdi are present under RemoteInstall, and check the relevant network paths and TFTP settings. An oversized TFTP block or window can cause timeouts, so test a smaller setting if the trace supports it.
Also record the client’s exact PXE error and firmware mode. Your separate DHCP server can still require the existing IP helper. Re-enable any security controls disabled for testing.