[pve-devel] [RFC manager v2] ceph: Allow specifying cluster network separately

[Thomas Lamprecht]

On 08/30/2017 01:17 PM, Philip Abernethy wrote:
> We always recommend putting ceph-internal communication into a separate
> cluster network. Yet pveceph init does not allow to specify one,
> leading to difficulties if one were to follow the setup guide on the
> wiki, which proceeds to create monitors after the ceph network was
> initialised.
> ---
> That looks a lot cleaner and is easier to understand. Thanks for the suggestion
> I didn't see the man-page update accordingly. Is that a manual step?
> 

Yes. man-pages are generated from the pve-docs repo, using the pve-doc-geneator
deb package. But the final manpage is then still shipped in the respective package.
So such a update cycle normally consists of:

* install the changed package
* switch to pve-docs run `make update` and then install this updated package
   (make clean && make dinstall)
* re-install your modified package from step one and done.

This can be shortened by using the $PERL5LIB environment variable, with this
the first step could be skipped.

* export PERL5LIB=<PATH-TO-REPOs-PVE-DIRECTORY> (note: the path should not
   include PVE/ but end a level above) and then make update and make dinstall
   in pve-docs. This allows picking up description/schema changes from the
   development repo directly.
*  make dinstall in respective changed package