Handful of clients reporting Policy failure (yet still installing policy)

[Cyrus11]

I've got 3-4 clients that will install policies but fail somewhere in recon after install. When I run recon from the command line, this is what I get.

bash-3.2$ sudo jamf checkJSSConnection
Password:
Checking availability of https://casper.server.com/...
The JSS is available.
bash-3.2$ sudo jamf recon
Retrieving inventory preferences from https://casper.server.com/...
Finding extension attributes...
Locating hard drive information...
Locating hardware information (Mac OS X 10.8.5)...
Locating accounts...
Locating applications...
Searching path: /Applications
Locating package receipts...
Gathering application usage information...
Submitting data to https://casper.server.com/...

There was an error.

     Message has no content

bash-3.2$

I've reimaged, unmanaged and ran the -removeFramework command and then readded the machine and it still behaves the same way. The rest of the clients are seeming to behave fine, which is why I think it's the clients rather than the JSS. Has anyone seen something like this and have any advice on what else to check on?

[haircut]

"Message has no content" means something is tripping up the jamf binary. It sends a malformed XML request to the JSS, which it can't process so it returns an error. The JSS should return the ID of the computer object on success. Open a ticket with support and they will help you track it down.

My guy did some absolute wizardry to find an invisible unicode character in a couple plug-in paths. In my case, it was plug-ins for the very first version of InDesign CS. Specifically, U+0003 or the "end of text" character. Same character on three clients.

On the problematic clients, you can run the following command to create a reconForm file:

sudo jamf recon -saveFormTo /path/to/output/directory

Open up that {hostname}.reconForm file as plain text in a text editor, and search for the caret character (^). It may point you in the right direction. You can also try searching for that specific character I encountered by going to http://unicode-table.com/en/#0003 and copying the character, then searching in the .reconForm.

From there, I just deleted the plug-in since it was no longer necessary in my environment. Clients immediately started running recon properly.

[Cyrus11]

This seems like the type of shenanigans I'm looking for, thanks for the lead. Searched the reconForm and didn't find that particular issue but this gives me something to chew on. Thanks again, I'll post back when I makes some sense of this.

[Cyrus11]

Talked to support and the rep did initially think it was the unicode issue but even with the 9.21 update that fixes that the error persists and the relevant portions of the server log are spitting out...

[org.xml.sax.SAXParseException; lineNumber: 0; columnNumber: 0; cvc-complex-type.2.4.a: Invalid content was found starting with element 'application'. One of '{device}' is expected.]

This is already apart of an existing defect D-005738 of which they are aware, hooray. I can now sleep better at night even if I don't have a satisfying fully green pie chart on my dashboard.

[bsuggett]

Something we've noticed is that if you perform
sudo jamf removeFramework
then
re-enroll the machine the problem seems to go away