Odd network problem with InstaDMG-generated image

[Anthony Reimer]

Since September, we have had an odd problem in our labs where a user will sit down to a machine which is already logged in (using a local, standard account with some MCX controls), will open (or switch to) a web browser and not be able to surf the net. Cached pages may load, but new ones will not. I can access all other networking functions perfectly, including file sharing and ARD remote control.

The fix: logout then log back in to the account. The problem does return eventually, but logging out and back in (or using Fast User Switching to login to the admin account, open a web browser — which then works — logout of the admin account and log back in to the standard user account) is the immediate fix.

This problem happens only periodically, but enough to be a PITA. It does not seem to be model-specific (most of our Macs have had the problem at one time or another).

The images being deployed are both a pure InstaDMG image and an image that was generated by InstaDMG then manually tweaked and cloned. They have been deployed both by DeployStudio (1.6.6) and Disk Utility.

We zapped the PRAM on all the machines yesterday just in case that might fix it. No idea if it has worked (too soon to tell).

Anyone else having this problem? Anyone have any suggestions about what files in the master might need to be killed? (I’m suspecting a rogue package, probably one that I created using Composer or PackageMaker.)

[u]Tech specs:[/u]
Mac OS X 10.5.8 (10.5.0 disk image with combo update package)
InstaDMG 1.5rc1
Two local accounts: one admin, one standard (with MCX settings added after the fact)
No network accounts
Images built on a Mac Pro (early 2008)
Deployed to Power Mac G5 (all generations), Mac Pro (2006) and Mac mini (late 2009) machines.

[Anthony Reimer]

Just an update: zapping the PRAM seems to have solved the problem. No recurrences for a week now. 🙂

[Anthony Reimer]

I spoke too soon. The problem resurfaced on three different machines today. Any ideas — even troubleshooting — would be appreciated.

[knowmad]

I had a completely unrelated issue on a windows machine yesterday that I will retell here because it might help…. bear with me.
A user had her machine looked at by one of our developers because she could not get online in the office using wireless.
Wired worked fine.
The developer could not fix it.
The next day the issue disapeered in the office, but surfaced at home.
Full connection, strong signal, working IP address, could ping the router/gateway/other machines….. only cached web pages.

worked on it over the phone and on a whim, did a team viewer session with her and it worked…. I got into the machine, but it still could not serf the web wirelessly…. wired worked fine.

I did an nslookup and discovered that her DNS had been hardcoded to that of our office.
It appears the developer when trying to help her hard coded the office dns into her wireless and when he put the machine back onto DHCP, he forgot to clear that.

I cleared it, she worked.

SO:
1) I assume these are desktops, but do they have more than one connection?
2) have you checked DNS on all connections?
3) IP conflicts?
4) log files?
5) little snitch?

recently we also had a bad IP address… which is to say a perfectly normal usable address being handed out by DHCP that stopped any machine that got it from working properly (had to do with another developer but we wont go there). Any chance all the bad instances have the same IP?

[Anthony Reimer]

Thanks for the reply, knowmad. Sorry for the slow response.

I just realized that I left out one important detail: all machines are using fixed IP addresses (“Configure: Manually”). I have two DNS servers listed in the configuration file (the same two that show up in grey when other local users connect via DHCP). The machines are all desktops (Mac Pro, Power Mac G5, Mac mini) and are connected via Ethernet. They are all on the same subnet. Most are on the same physical switch.

The other possible issue is that I have had to use a hybrid approach, where InstaDMG generates most of the image, but because of time and staffing constraints (I am a one-man show for the most part), I do take that image, install it on a test machine, tweak a few things manually and then clone that for deployment. The network settings are one of the things that gets tweaked — essentially, they all get the same fixed IP address and then we manually change it on first boot (after login). Now that I have had success getting DeployStudio set up, I am going to abandon this tweak and use DS to set it, but I haven’t recreated my image this way yet.

The log files that I have seen have not provided helpful info, but it is rare to catch this problem at the time of failure. Earlier this week, we did have a failure when the user was in the middle of a session with a web-based e-mail client, but that’s the first time I have lucked out in that way.

More ideas welcome (as I try to build a simplified image for long-term testing). Thanks.

[Author]

Posts