Marc A. Volovic wrote:
> I also heartily recommend AGAINST running a pure-raid system. At
least the
> swap partition should reside on a non-raid device (else you raid virtual
> memory, an exercise in futility and waste).
If the disk fails during a swap-in (or swap-out) does my system crash ?
does the process whose page it is crash ? does the swapfile on the
broken disk get forcably removed and thus all processes who have pages
in it get killed (via SIGBUS or something). Which could be a lot of
processes and leave your remote system without a working SSH or
something just as important as it.
If my system crashes, I'll keep my swap partition(s) RAIDed. This
options seems safest to me.
If processes crashes then it depends what your High Availability
requirements are for the system in question.
The dynamics of swapfile requirements these days is different, most
systems have ample RAM to use for OS disk buffers rather than running
their main workload through a swapfile (as it was 15 year ago) people
just buy more RAM as soon as their system needs it.
If anyone has any factual information to add to how Linux handles this
situation, I'd like to hear.
Thanks
--
Darryl L. Miles
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
