Various wrote:
> > I MUST be missing something simple - has nobody else seen this?
>
> Ian Darwin has a laptop that is mostly busy handling an interrupt
> storm.
Yup. Lots of possibilities. Before spending too much time diagnosing
any one theory, you need to come up with a way to triage theories.
Some possible causes:
slow cpu -- wrong clock speed?
broken cache? bizarre bus problem?
interrupt load
serial? network? something else?
disk subsystem
many. ata dma is a common culprit.
but a ufs filesystem nearly out of space
will perform pretty badly even on good hw.
So some early experiments you should try:
vmstat 10
On an idle system: should report 99-100% idle,
low counts on interrupts, no paging, etc. Run on
both bad & good system, compare results.
time openssl speed rc4
this is a 100% cpu-bound case.
If it doesn't come up with similar numbers,
you know it's not a disk problem. If the user
time isn't nearly equal to elapsed time,
you know something is stealing cpu, otherwise,
you know the cpu is doing something bizarre.
time dd </dev/rwd0c >/dev/null bs=100k count=1000
this should be fast, user & system time should
be low, etc. Significant system time means
you should investigate dma options on your
local disk.
time tar xf ...
(unpack something new; don't unpack over something
that exists; you want to allocate directory entries, inodes,
and filespace.)
this will catch ufs weirdness.
Once you know which of these has problems, then you can start
chasing down that branch.
-Marcus Watts
