PasswordService in Uninterruptible wait

[mosx86]

Is it normal behavior for the PasswordService to go into an “Uninterruptible wait” status intermittently on an OD master? Top was showing a stuck process so after investigating with ps aux, I have found that it’s the PasswordService that keeps going in and out of the stuck state.

According to the document [url=http://unix.derkeiler.com/Newsgroups/comp.unix.programmer/2005-08/0313.html]here[/url], a stuck process is waiting for a kernal function to complete and can be caused by an incomplete I/O. I do notice that my disk is pretty active:

% iostat -w 1 disk0
disk0 cpu
KB/t tps MB/s us sy id
8.37 1371 11.21 32 8 60
8.22 543 4.36 19 3 78
8.24 1751 14.08 12 12 77
8.18 420 3.35 18 4 78
7.50 16 0.12 18 2 80
9.72 1439 13.67 36 8 56
8.22 2691 21.59 27 15 58
8.22 2901 23.29 22 10 68
11.04 779 8.40 44 11 46

The same is true on my replica:

% iostat -w 1 disk0
disk0 cpu
KB/t tps MB/s us sy id
8.48 208 1.73 1 2 98
8.23 2020 16.24 1 6 94
8.22 1756 14.09 17 10 74
9.36 2635 24.08 6 17 77
8.23 3251 26.13 2 12 85
8.24 3226 25.95 4 13 84
8.24 2372 19.08 2 12 86
8.57 1945 16.28 2 14 84
9.94 1262 12.25 22 4 74

Both seem kind of high to me, even with the I/O associated w/ lookups…

[mosx86]

[QUOTE][u]Quote by: MacTroll[/u][p]1. are you reacting to something not working or just poking around in top and seeing the stuck process?[/quote]

Over the past few weeks users have been complaining of their systems suddenly hanging for a few seconds intermittently, including hanging up to minute or so on logins (we’re on network homes). I’m not seeing any abnormal traffic on the either the network or from the file servers. Users aren’t getting any errors that they’re losing connectivity, everything just seems to hang and then go back to normal. I’ve looked in the AFP logs, the system log, the asl log, the DS logs and am not finding anything unusual.

I was looking at top when I noticed that PasswordService was “stuck” and I don’t recall ever seeing it in this state. I don’t think PasswordService would be causing the problem but am wondering if could be a problem with DirectorySerivce.

We ran into a little problem a few weeks back when I updated the Master and Replica to 10.4.11 and the replica decided it was the parent and borked our password database. I’ve worked w/ Apple on that and ended up doing a deep clean of the replica before recreating it. The hanging issue matches the timeline for the problems we had with the replica.

[quote]2. What else is your system doing?[/quote]

The system is an OD master and also runs our PDC. We have one replica in the loop. So aside from handling those duties it’s not doing anything else. Both master and replica are running 10.4.11.

[quote]If you do have a lot of I/O it’s not entirely out of the question that PWS would be waiting on a disk operation to complete before it could continue.[/p][/QUOTE]

The master seems to get the lion’s share of lookups, I’m wondering if the disk activity is being driven by this, though it still seems a bit high to me.

[mosx86]

Still trying to track down this issue where clients will momentarily hang. I suspect it to be an issue with DirectoryServices as in a few cases, users have lost access to their files while they are still mounted. Restarting DirectoryServices clears this up.

What I’m not sure about is how to track this down to a problem on the clients or is it merely symptoms of a larger problem with my OD/replica setup.

Any ideas?

[Author]

Posts