[vox-tech] Three Install Questions

[Rick Moen]

Quoting Peter Jay Salzman (p at dirac.org):

> 1. Having multiple swaps per disk is of almost no benefit at all.

So, a client of mine builds Linux servers, and provides a custom RHEL
3.0 build image for one particular customer that buys a lot of 4 GB RAM
dual Opteron systems with dual 73GB SCSI drives.  Customer management,
for reasons that passeth understanding, insisted they wanted 32GB of
swap space on each node, and moreover wanted all of it in a single
partition on the boot drive.  Such systems were duly delivered, but
customer soon reported that the systems were hanging hard while in
production use.

Extensive tests followed using the Cerberus Test Control Suite (CTCS),
which stress-tests Linux boxes using a number of simultaneous software
processes including iterative kernel compiles.  With a single 32GB swap
partition, CTCS induced a system hang in 1 day.  With a pair of 16GB
swap partitions, CTCS hung the box in 2 days.  With four 8 GB swap
partitions, five days.  And thus up to the ideal configuration, using as
many 2 GB swap partitions as the limits on SCSI device numbers
permitted, in which, as far as we could tell, CTCS ran forever.  The
customer's load image was duly modified, and the cause has been
presumptively attributed to hitting a bug in the RHEL 3.0 kernel's VM code.

(Yes, there was indeed no rational reason why customer needed to deploy
all that swap in the first place.  But that's not the point I was
making.)