RE: [SSI-users] Migrated processes - what if the node dies?

[Walker, B. J <bru...@hp...>]

Stateful restart is very application dependent and may or may not affect
performance significantly.  One has to understand the application and
make to code changes to save state at the appropriate times and then be
able to recover that state in a restart to get "back to where you were".
For some applications it may be next to impossible to do this feasibly.
For init it was pretty easy.

Bruce


> -----Original Message-----
> From: Chris de Vidal [mailto:ch...@de...]=20
> Sent: Monday, March 01, 2004 12:10 PM
> To: Walker, Bruce J
> Cc: ssi...@li...
> Subject: RE: [SSI-users] Migrated processes - what if the node dies?
>=20
>=20
> Walker, Bruce J said:
> > C. stateful restart - this is what we used for init(8). =20
> The modified
> > program/daemon writes it state to disk so that if it is=20
> restarted, it can
> > read the state in and hit the ground running.  The shared root
> filesystem is
> > a useful place to write the state since you know it must be=20
> around at all
> > times.
>=20
> Ahh, I was having a hard time understanding the=20
> chicken-and-egg problem
> that init poses :-)  Very cool.
>=20
> Is that easy to apply to any process I want?  How does it perform?
>=20
> /dev/idal
>=20