Thanks to all who responded. The overwhelming majority view
(10 out of over a dozen responses) supports mirroring the swap
device when the root disk is mirrored. The reasons range from
the fact that it doesn't cost much, can't hurt and
[will definitely | should | might] (pick one) avoid a system
panic. I've intentionally induced kernel panics by cat'ing
/dev/null > /dev/swap; this to test a failover system. I
recently tried it with a new E-250 and nothing happened. But
the system was idle.
Real life isn't usually so neat, as email@example.com points
Murphys law states that if its gonna happen, it will happen at your
bussiest time of the year when the machine is pegged and you are home
sleeping in the dead of winter.. What could mirroring swap hurt? I do.
Its nice to know that if a disk fails the machine will definatly keep
running:) I get enough after hours calls for stupid stuff as it is.
I dont need real problems at 3am.
One minority view characterizes mirroring swap as "superfluous
overhead," but being in a shop that takes the other position
on this point of doctrine, this person holds that view quietly,
to avoid being charged with heresy so I will withhold the name.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:34 CDT