:I had last night a panic on the console with the CVS checkout of last
:Thursday (no VFS changes from DragonFly-Stable).
:
:Unfortunately I failed to make an exact note of the traceback this
:time but it was a "vfs_cache" error"with something about malloc being
:exhausted. One disk was unaffected but the another seemed to lose its
:disklabel with "no magic" on partition errors.
:
:In hindsight I wish I had saved the disklabels for the disks (which I
:have done now!)
:
:If the problem happens again I will send a traceback.
:
:Cheers Steve
Run the following:
vmstat -m
sysctl kern debug | fgrep vnode
sysctl hw.physmem
sysctl hw.usermem
df
disklabel blah ... disklabel output for all your disks
disklabel blah
disklabel blah
Any sysctl or boot.config tweaks you might be using.
The VFS cache shouldn't run out of kernel malloc space, if it is
there could be a leak.
The blown disklabel is more worrysome. If it occurs again be sure to
save the first, oh 64 blocks or so of the dead slice so we can examine it
post-mortem. This is the only recent report of disklabel corruption
so make sure it isn't cockpit trouble somewhere (well, if that's possible
now that you've recreated the label :-)). e.g. check for overlapping
partitions and if you have kernel dumps enabled check that you are
dumping to the correct partition (usually the swap partition), and that
it is big enough to hold the dump.
-Matt
Matthew Dillon
<dillon@backplane.com>