Broken mds and Spotlight...

Question

Hi,
I tried using my AppleCare contract to try to get to the bottom of the
"mds keeps crashing and restarting the index process over and over"
problem that occured after I upgraded to SL.
They collected all of the crash dumps and after a week the report
that I got back was that there 'evidently' a problem with (at least) one
of my "user" files that was causing the mds crash.
There might be some truth to this...
I restarted the indexing with just the "System" directory and when indexing
was complete I added /Library waited, added /Applications waited, and then
Developer and waited for the indexing to finish...
I have two users Myself and Postgres.. the postgres account is only used for
compiling the postgresql database...
Adding the home directory for either of the two above users causes mds
to fail the error as logged in the system messages file looks like:
Oct 12 09:33:35 mbp mdworker32[2387]: kCGErrorFailure: Set a breakpoint @ CGErrorBreakpoint() to catch errors as they are logged.
Oct 12 09:34:38 mbp mds[2377]: /SourceCache/Spotlight/Spotlight-507.1/index/ContentIndex/CIPositions.h:215: failed assertion 'len == 0 || bytes[len-1]==0' /.Spotlight-V100/Store-V1/Stores/548E69A1-D43A-4775-82ED-9B177899831E (hfs, t: 0x11, st: 0x1, f: 0x480d000)
Oct 12 09:34:40 mbp com.apple.ReportCrash.Root[2412]: 2009-10-12 09:34:40.583 ReportCrash[2412:2903] Saved crash report for mds[2377] version ??? (???) to /Library/Logs/DiagnosticReports/mds2009-10-12-093440localhost.crash
Oct 12 09:34:42 mbp com.apple.launchd[1] (com.apple.metadata.mds[2377]): Job appears to have crashed: Segmentation fault
Oct 12 09:36:22 mbp mdworker32[2425]: kCGErrorFailure: Set a breakpoint @ CGErrorBreakpoint() to catch errors as they are logged.
Oct 12 09:37:12 mbp mds[2413]: /SourceCache/Spotlight/Spotlight-507.1/index/ContentIndex/CIPositions.h:215: failed assertion 'len == 0 || bytes[len-1]==0' /.Spotlight-V100/Store-V1/Stores/548E69A1-D43A-4775-82ED-9B177899831E (hfs, t: 0x11, st: 0x1, f: 0x480d000)
Oct 12 09:37:13 mbp com.apple.ReportCrash.Root[2437]: 2009-10-12 09:37:13.400 ReportCrash[2437:2903] Saved crash report for mds[2413] version ??? (???) to /Library/Logs/DiagnosticReports/mds2009-10-12-093713localhost.crash
Oct 12 09:37:13 mbp com.apple.launchd[1] (com.apple.metadata.mds[2413]): Job appears to have crashed: Segmentation fault
Oct 12 09:37:13 mbp mds[2438]: (/)(Error) IndexGeneral in SIOpenIndex:checkcrashstate: 1 for /.Spotlight-V100/Store-V1/Stores/548E69A1-D43A-4775-82ED-9B177899831E
Oct 12 09:37:13 mbp mds[2438]: (Normal) DiskStore: Rebuilding index for /
Oct 12 09:37:13 mbp mds[2438]: (Normal) DiskStore: Creating index for /
The above will repeat as long as I allow indexing...
Each of the directories contain thousands of files and many directories I cannot see how
to find the files that cause the crash.
I might add that I did not have this problem with Leopard.
Jerry

Answer 1

548E69A1-D43A-4775-82ED-9B177899831E
This is the same UUID as before. My take on this problem is that Spotlight is trying to index some disk but something is wrong on that disk which makes it crash. As a first step we can try and figure out which disk is causing the problem. You can do this:
Open the terminal and type this command:
mount
This should give and output like this:
/dev/disk0s2 on / (hfs, local, journaled)
map -hosts on /net (autofs, nosuid, automounted, nobrowse)
This will give you a list of all the mounted filesystem. Start with the hfs volumes. The part after the on is what you need for the next step (/ in the first line in the output above and /net in the second line but the second one isn't hfs.) Now run this command:
diskutil info / | grep UUID
Replace the / with each line of the output that you get until you find which disk has the UUID in the console message.
(Sorry this is the only way I know how to find the disk for now.)
The next step would be to verify that disk through Disk Utility if there are any errors and lastly to locate the 'bad' file. I hope that you succeed with that.
Cheers.