Closed GoogleCodeExporter closed 9 years ago
If I had to change the certificate of one of the virtual servers, I'd most
probably use 'Graceful Restart' after I either
changed the file or the configuration.
This RFE doesn't make much sense to me. Stefan, could you please elaborate on
the rational behind the request?
:-?
Original comment by alobbs
on 17 Nov 2008 at 9:13
Because any graceful restart now will just kill the server and start a new
instance,
hence ask for the PEM key of the certificate again (at the console) thus it is
no
graceful restart, it actually kills the non HTTPS instance too.
Original comment by ste...@konink.de
on 17 Nov 2008 at 9:15
Yeah, it would be a problem, actually. However, I don't know what the best
solution would be.
While reloading the certificate is desirable, re-entering the password is not..
and that's actually a problem.
I'm not sure whether we'll be able to fix this issue without implementing some
sort of mechanism with which
cherokee-admin could ask for the certificate password.
Ideas?
Original comment by alobbs
on 17 Nov 2008 at 9:41
Firstly the http part should *always* be started. It should not block on the
HTTPS
part. Secondly I think the PEM part is process specific; now I don't want to you
reimplement the cherokee reloading completely because I understand how it works.
I wonder; this could get into real issues btw; imagine what happens if a https
capable worker is restarted due to a problem. It would block the entire server.
It
might be good to give the parent process the certificate.
Original comment by ste...@konink.de
on 17 Nov 2008 at 9:50
Yeah, that makes sense. Cherokee could store the passwords and pass them to
cherokee-work each time it's
restarted.
I'm not sure about the technical side, though. What would be the way of reading
and passing the passwords of
multiple certificates? It sounds kinda messy to me. :-m
Original comment by alobbs
on 17 Nov 2008 at 10:10
If you want to mark Cherokee as security leak where it is possible to read out
passwords from memory. I wouldn't take that approach ;)
But storing the *actual* extracted certificates. Sounds reasonable? Since that
happens always anyhow.
Original comment by ste...@konink.de
on 17 Nov 2008 at 10:21
I would just suggest to take the easy approach and use passwordless root-only
certs.
Closing this bug.
Original comment by ste...@konink.de
on 10 Dec 2008 at 10:16
Original issue reported on code.google.com by
ste...@konink.de
on 14 Nov 2008 at 1:14