> Once the "vos release" is actioned the client's next chunk request to
> the server will get a chunk of the new binary. I think that mixing
> pages of the old and new binaries won't work and will have unpredictable
> results (eg X11 freezes or program core dumps).
>
> Does this make sense? Can anyone else verify this?
> If it is so then what is the best strategy for doing "vos release"?
Yes, perfect sense. Processes have terrible problems if their binaries
change on a vos release. If the process in question is running as root,
this will quite probably crash the client as garbage is liberally
sprinkled round VM. We've seen this on Ultrix, and the best response is
- on updating a critical binary, stop all running processes using it
during the vos release.
Now, can you explain to me why the X server hangs on a Sun even when
the binary HASN'T been changed? We frequently vos release the volumes
containing Ultrix X binaries, and I have yet to see an X server hang as
a result. Could this be the one advantage of not having shareable
libraries? It's difficult to know what other significant difference there is.
Peter Lister [EMAIL PROTECTED]
Computer Centre,
Cranfield Institute of Technology, Voice: +44 234 754200 ext 2828
Cranfield, Bedfordshire MK43 0AL UK Fax: +44 234 750875
