Say you have a site that is connected to the data center with low-latency fiber-no VPN. Now, say the use a DFS Namespace link for all their files. Users are mostly engineers creating run programs for automated cutters and users loading programs. They have been happily using this since they were acquired more than a year ago. And then, ALL OF A SUDDEN, their folder structure disappears! Only item left was the folder that used to contain their roaming profiles (after acquisition, we discontinued roaming profiles and copied their folder structure). Worse yet, it happens intermittently to many diverse users. Some with new engineering class PCs, some with laptops on decent wireless connections, and in between. Seems to be no way to reproduce itself and fixes itself after a few minutes.
OK, so we went through the normal troubleshooting process. Doesn’t appear to be the network, they don’t lose connection to other resources (Citrix, etc.). Flood pinging and load simulation doesn’t reproduce the issue.
We found DNS wasn’t updating for this scope. OK, we upgraded the server and failed to put in the service account for secure updates. Didn’t fix it, but it made us feel like we did for awhile. Did I say this could not occur for a few hours then magically come back?
Spent several days going through our DFS environment with Microsoft Premier Support. Found some things that could be cleaned up, but no smoking guns.
You know MS is at a loss when they assign you the homework of grabbing a screenshot, clearing the caches, and forwarding them a trace. We got that to MS, and they wanted to be sure the problem was actually happening in the trace because it looked “normal.”
At this point, we finally stepped back. This should have happened earlier. What do we know? It only happens at one site. I am not able to reproduce it on demand.
Well as some of you may know (we didn’t), you should never use DFS-N and offline files. At first, not a good clue when I read this. We do not use offline files, thankfully. However, this was an acquisition. When I explored their files system that a previous engineer had just dropped on our file system, I discovered there were offline files!
Turns out, they used roaming profiles previous to us integrating. Roaming profiles file structure is “available offline.” Once this was turned off, the issue disappeared immediately.
I have linked some of the technical articles below if you want to read why this doesn’t work, but the simple story is that DFS-N addresses are stored as server paths in offline and not links. The users clients were experiencing periodic short spikes in latency do to less than optimal network equipment in the building. This was causing a “feature” of Windows 7 to kick in. In Windows 7, slow link mode is enabled by deafult when file server to client communications exceed 80 ms. I was not able to find out how it measures this metric, but it was clear we were periodically exceeding it. The default is to recheck this metric after 2 minutes-that’s when our files would repopulate seemingly random.