[profmathers]

You are officially the first person I’ve met with the same problem. Somehow, I feel less like I’m the only guy in Ohio with a sunburn in January. What, if anything have you tried to resolve the issue?

M

[profmathers]

waaay repeatable, even somewhat documented. no crashes. i have a bunch of dated text files of users that are missing large portions of the numbered users (almost all of them have all of the faculty users) from the last few days. the problem’s been happening for a while, but it’s worsening gradually. ironically, all users can log in, all the time. no end-user complaints.

of course, the one time i manage to nail my company’s other linux admin down long enough to look at it…it exported almost 900 users. don’t i look the fool?

ah well, it’s friday. one guinness down, several to go! have a good weekend!

[profmathers]

just ran the same command again, to demo the problem to a friend…and got the whole user list. ARGH!

M

[profmathers]

Woo…good ideas, and I’ve got a scary result:

No, I don’t have the user list capped in WGM (and this happens on WGM from multiple machines including locally on the server.) I did run a series of queries in dscl:

running this command:

# dscl -u -p /LDAPv3/127.0.0.1/ -list /Users -> /path/to/outputfile.txt

I received different output each time. None of the output listed all users, in fact less than 2/3 each time.

Running dscl queries on individual known users however, returns correct information.

Here’s another odd piece of info: the school Tech Director (I am a consultant) used what I’d call an unusual shortname scheme when she imported her users. Each student has the last two numbers of their graduating year as the first two digits of their shortname, e.g., 05rennichj. There are between 150-300 users in each class, meaning that the database has to parse a kazillion* records, each with the same first character and a similar second character. The only reason I let her do this was because I had already spent half a day convincing her that she couldn’t use the last 4 of a user’s SSN for UIDs, as they’re not unique. Yes, I did have two users whose SSNs ended in 0501. It’s funnier now than it was then.

The reason I thought this funky shortname business might be relevant is that the display of faculy, staff and admin accounts which do not have the numeric prefix is much more reliable–my output text files from dscl are about half faculty accounts, where sometimes entire classes of students are not present.

Thanks so much for your quick reply–hopefully it’s not just a blown-up OD database…but I’m running out of red herring

M

[Author]

Posts