[email protected] wrote on 02/13/2014 11:34:56: > On Thu, Feb 13, 2014 at 09:38:53 -0500, > [email protected] wrote: > >I just had interesting (or incredibly stupid) idea concerning the critical > >limitation of the live overlays where the host crashes with bus errors > >when the temporary space for the COW is exhausted. Is it possible to use > >trim/discard support to mitigate some of the unidirectional growth for > >when files are deleted and/or overwritten? Conceptually it seems like it > >is a good fit, but I've not investigated the practicality. Just curious > >if this has been explored or if there are easily known reasons why this > >wouldn't work. > > I don't think this will help. You aren't going to recover space from the
> base file system, because it never changes. As more an more packages > are updated you are going to use more space in the overlay area. I figured as much otherwise somebody would have already done this. But my web searches didn't seem to reveal anything so I had to ask. > Portable flash drives are getting cheaper and you might consider getting > one large enough for a normal install and use that instead of a live image. > Depending on how much you are installing, a 64 GB flash drive is probably > large enough. That won't help me where I'm looking at hundreds to thousands of embedded nodes that have very specific use requirements for being based on a live image. Using a live image makes them very resilient to mutation/corruption and anybody can simply reset the hardware and have a node back to its normal working state. You can think of this setup behaving much like firmware, albeit with a Fedora image. :-) I just have to ensure the workloads can never cause the overlay to become exhausted, which is sometimes easier said than done, which is why I asked. -- John Florian
-- livecd mailing list [email protected] https://admin.fedoraproject.org/mailman/listinfo/livecd
