Memory issue when running Casper Imaging

[sburrows]

Anyone else experiencing memory issues on their server when running Casper Imaging? We are using a mid 2012 Mac Mini with OS X with 16 GB of RAM. Casper is at 9.31, OS X 10.9.3 and server app at 3.1.2. When imaging a machine the memory usage on the server skyrockets to 15.99GB and imaging comes to a huge slow down.

We have noticed this issue when the server was running 10.8.5. We upgraded to Mavericks Server and changed the distribution point to use SMB. This seemed to resolve the issue for awhile, but now it's back.

[bentoms]

I'm not seeing this at all.

JDS or AFP distribution? On the same server?

[sburrows]

Was AFP distribution then we switched to SMB with better results. The one server is doing all things Casper. We do have some other server roles enabled on it.

[sburrows]

Just noticed it is also happening when I was replicating distribution points in Casper Admin as well. It leads me to believe there is something going on with file sharing over both AFP and SMB. I'm going to try transferring a large file over from the server and see if that causes the high memory usage as well.

Update: SMB transfers natively work fine, so it seems like we are having some sort of issue with AFP/SMB transfers in Casper Suite. I have a case open with support. Just waiting on the next steps they want me to take.

[were_wulff]

Hi @Burrows

Generally, this does tend to be an issue with AFP and OS X; it seemed to be much more prevalent in 10.7 and 10.8, but we do still see it in Mavericks now and again.
Looking through Apple's discussion forums, slow transfers & high memory usage with AFP appears to have been an issue since 2006.

I think I may have e-mailed you on the case that's currently open earlier today, but just in case it gets missed, there is another thread that details similar issues along with some things we tried that seemed to help in those cases: https://jamfnation.jamfsoftware.com/discussion.html?id=10171
It's worth a read, as there didn't seem to be one solution that worked 100% for everyone; sometimes it had to be tweaked a little to get it going in some environments.

One thing we found, in a few instances, was that just setting the distribution point to SMB in the JSS didn't seem to be enough and we had to go into Server.app and specifically uncheck "Share over AFP" for any file shares that we had set up that the JSS would be using, or the behavior (while it might lessen) was still present to a noticeable degree.

In one case, we also ran into the issue where, if the distribution point and JSS were on the same computer, we'd see AFP file shares be incredibly slow, not release connections, or chew up memory and CPU, but if the distribution point was moved to a different drive partition or to a different computer/VM entirely, the behavior seemed stop.

Amanda Wulff
JAMF Software Support

[sburrows]

Thanks. Yes I received your email. I will be testing out your suggestions shortly.

[bentoms]

@Burrows, FWIW we never image from our master DP & never sync whilst imaging.

So that could be why we've not seen it.