Home Forums Mac Software Mac Operating Systems Your startup disk is almost full

coming up on an imac here running 10.5. an imac with 67gb free on the hard drive. After having run diskwarrior and safe booting.

ever seen this before?

furthermore, activity monitor shows a continuous 17mb/sec disk io. (read and write, very close in sync, almost as if copying, but jagged, not smooth) and mds is of course ruling the roost at 33% cpu. It's been like that for the last EIGHT hours. Machine is extremely sluggish to respond. (this is an imac g5, it can't really take this level of abuse in stride) Guessing the disk io, mds cpu,l and erroneous lowspace warnings are related.

I feel silly suggesting this to you because I'm sure that you already know this stuff but, just in case, read Where Did My Disk Space Go by MMT3.

an imac with 67gb free on the hard drive.

If ooonly it were that simple...

Well this is what my iMac says under Disk Utility

Mount Point : /
Capacity : 232.6 GB (249,715,376,128 Bytes)
Format : Mac OS Extended (Journaled)
Available : 218.6 GB (234,755,248,128 Bytes)
Used : 13.9 GB (14,960,128,000 Bytes)

Er.....some joke I'm missing?

Well this is what my iMac says under Disk Utility

what I mean is, clearly that warning I'm seeing is not supposed to be coming up.

Warning from what/where/who?
Is any Microsoft program (Word, etc) running perhaps?

Have you slapped the volume with DiskWarrior yet?
oops, skip that one.

Hmm,

Yeah... need to get to the bottom of that. Sounds like indexing has encountered an "overly-interesting" file. Try using lsof and fs_usage to see if that's the case.

sudo lsof -c md # looks at all procs starting with md

sudo fs_usage -f filesys mdworker mdimport mds

[you can knock off various processes from that last one. sometimes it's helpful to see what only one of those is doing.]

Are we sure that mdworker and/or mdimport are even operational?

ps axcru |sed '1p;/ md/!d'

If mdworker is absent, the indexing is endless.

sorry I didn't specify, I am referring to the Mac OS X warning dialog that normally pops up when your startup volume drops below about 500mb of free space.

A&I appears to have been required. Looking good so far. Can't have been a spotlight database problem or this would not have fixed it.

Not the typical type of Spotlight problem, no.
Was mdworker running or not? Didn't check?
Too bad... now <nobody> will know for sure.

mdworker was not running. mds was, and was eating about 1/3 CPU. I suspect it was also the cause of the ~17mb/sec read AND write disk io, and as a side effect of those two, what was thrashing the machine so badly.

The weirdest part of it all was that the disk io was somewhat jagged along the top, not what you'd expect say when copying a file or verifying a disk image etc where it would be more or less flat along the top, AND the read & write were both in perfect sync, it was hard to spot the red of the write under the green of the read as they so rarely diverged. I can only assume this is a file being duplicated.

Lookie Here

Hmm, haven't done an A&I in ages. Any chance there are still any system.log files laying around, from the time period when the problem existed?

Entries such as these would resemble that MacNN thread:

  mds [3157] (Error) Import: importer:0x901800 "Importer start failed"

  com.apple.launchd[1] (0x10f400.mdworker[299]) getpwuid("4294967294") failed

Most of the pids (other than launchd) would differ of course... but that "4294967294" uid value was key. It indicated a problem with uid -2, which turns out to be 'nobody'. (during an initial index, 10.5 has like 3 mdworkers going: one under root, one under the user and one under nobody). That thread was never 100% conclusive -- but from the procedures done and the clues left behind -- we surmised something must have been weird with /var/db/dslocal/nodes/Default/users/nobody.plist (which prevented launchd from being able to kickstart an instance of mdworker under nobody).

I looked in console the whole while it was doing it and didn't see anything useful going to log besides an HP driver crashing/relaunching every few minutes, which may be related.

I'm not certain, but i don't think those two log entries i previously posted would be spilling out continuously "the whole while" -- but rather appear only a few times (2 or 3), probably back a ways (shortly after either startup or login). To be sure, a search for either 'Importer start failed' or '4294967294' might be more conclusive.

I didn't see anything like that. the hp errors were only occurring every 3-5 minutes which is not what I was expecting.

Like i said, it's not about seeing... it's about searching.