My query related to large VM proceses hanging in D state, and sleeping on
kernelmap - since they are sleeping at a negative priority they of course
can't be killed.
Some suggestions that I and/or the net have come up with:
1. Hal Stern sent a complete answer (as usual), saying that sleeping on
kernelmap meant that the kernel was trying to allocate more memory from its
heap, and failing (heap full).
There are patches available to fix several kernel heap leak bugs. The patch
Hal suggested was 100330. However, that patch is superceded by 4.1.2 plus
patches 100516 and 100570.
Patch 100330 fixes many bugs, so the other two patches (for 4.1.2) may or
not be related to kernel heap leaks - I haven't got those patches yet so I
don't have the READMEs.
BTW, a regular reboot should clear the kernel heap, if you can afford such a
drastic measure.
2. There are bugs related to having a large amount of swap under 4.1.2; I saw
a thread in comp.sys.sun.admin a few months back reporting 690s with 4 CPUs
hanging under Oracle. This was explained by a Sun employee as being related to
the amount of time it takes the kernel to reorganise its notion of the swap free
list. Patch 100645 fixes this. In particular, it fixes the following:
a) Swapping to a UFS file (swapfile) can hang the system;
b) Large VM processes take a very long time to exit.
-- -------------------------------------------------------------------------- | Adrian Booth | | | Adrian Booth Computing Consultants | Ph/Fax: (09) 354 4936 | | 7 Glenrowan Pl., Willetton WA 6155 | abcc@DIALix.oz.au | | AUSTRALIA | | -------------------------------------------------------------------------- for a in past present future; do for b in clients employers associates relatives neighbours pets; do echo The opinions in this article in no way reflect the opinions of my $a $b. done;done
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:28 CDT