I’ll preface this by stating I’m not an expert at the different caching methods, I’m simply reporting what I’ve found with peer cache in my environment. YMMV
So as most companies are, in the trailing months of 2019, we’re trying to upgrade our remaining Windows 7 clients to Windows 10. We’re using a task sequence to accomplish this in ConfigMgr. The past couple weeks our upgrade numbers have slowed a bit, with some clients not even running the required task sequence. Hooray for more troubleshooting!
Let’s start at the basic level. I went to the Monitoring tab, Deployments, and looked at the required task sequence (there’s also an available one as well). I looked at the errors, and noticed the error “Program failed (download failed)”…What the fuck does that mean? At first I figured, maybe something wasn’t distributed to a remote DP. But it wasn’t until I got on a machine and saw the failures happen in real time, that I got an understanding of what’s going on.
Once I was on the computer, I initiated a retry from the software center on one of the problem PCs. It started downloading, I saw the folders in get created for the content in ccmcache, but after a short time, it failed again. I repeated this process a couple times while staring at the logs. Each time it failed, I got an error message in the execmgr.log that Content wasn’t available
Looking through the DataTransferService.log (ignore the timestamp, the log rolled over before I could get a screenshot), I could see the computer connecting to a peer cache source. I had started rolling out some super peers after returning from MMS Jazz Edition. What appeared to be happening is that the super peer was adding content to it’s local cache. I’m still unclear on how this works, but that’s what I was seeing. In the screenshot below you’ll see log files, specifically relating to the HP BiosConfigUtility. The log files aren’t native to the actual package itself. This presumably changes the hash of the package, and the client will no longer download that content.
Once I removed the Peer Cache policy from the list of machines, ran a machine policy on the super peer and local client, the client connected to the normal distribution points and the content started downloading. Again, YMMV, I just happened to come across this issue in my environment.