There is no existing connection, Win32 error = 53

A few weeks ago I saw that applications and packages was intermittently not being pushed over to our Secondary sites. When I looked in the sender.log I found the following:

There is no existing connection, Win32 error = 53
There is no existing connection, Win32 error = 53
Error during connection to \\<FQDN>\SMS_SITE (53). Error is considered fatal. Cannot connect to server <FQDN_of_server> at remote site <site_code>, won't try send requests going to site <site_code>  for an hour or until there are no active send requests.
Could not establish connection.
Attempt to connect failed.

Every time I saw the problem, I tried to restart the SMS_Executive service on the Primary site. Suddenly all queued up sending packages started to go out to the Secondary sites. Okey, problem solved? No, it would after a random time stop to work again, and the same error appeared.

My first suspicion was that the Primary site account was not added to the local group SMS_SiteToSiteConnection_<site code>. But why would it work after a restart of the SMS_Executive service, and then suddenly stop working? So I started looking into the permissions on C:\Program Files\Microsoft Configuration Manager\inboxes\despooler.box\receive to double check everything was in order. There I found out that theSMS_SiteToSiteConnection_<site code> group was indeed there. I added the Primary site computer account explicitly with Full Permissions for troubleshooting purposes. That didn’t do the trick, the problem still occurred intermittently.

When I opened up Windows Explorer and tried browsing to the \\<secondary-site>\SMS_SITE share, it worked out just fine. This was really giving me a headache, because I couldn’t really understand why it wouldn’t work.

At this point I started to see a pattern for my problem. Every time I initiated a new session to the share by either browsing or restarting the SMS_Executive service, it worked but then stopped working after a while. It hit me that we’re using WAN accelerators on our local office locations. Could they have something to do with it? I talked to our network team and asked them to let my Primary site to Secondary site traffic pass-through the WAN accelerator. When they had re-configured one of the WAN accelerators and I had restart the SMS_Executive service once again, it worked (as expected of course, since it worked every time after a service restart). But as a pleasant surprise, for that specific Secondary site with the re-configured WAN accelerator, it had after 2 days still not stopped working. So the network team re-configured the rest of our WAN accelerators, and to date it has still not stopped working. Woho!