[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Savannah-hackers-public] Setting up bzr+ssh on Savannah.

From: Sylvain Beucler
Subject: Re: [Savannah-hackers-public] Setting up bzr+ssh on Savannah.
Date: Tue, 4 May 2010 19:21:29 +0200
User-agent: Mutt/1.5.20 (2009-06-14)

On Tue, May 04, 2010 at 10:03:23AM -0700, Karl Fogel wrote:
> Sylvain Beucler <address@hidden> writes:
> >>> move the bzr service to the 'vcs-noshell' VM (instead of the current
> >>> 'sftp' VM) - so you don't disable SFTP access to the download area.
> >> 
> >> I didn't quite understand what you mean, sorry.  Can you expand it a
> >> little?  What exactly is the "download area" and how would it disable
> >> SFTP access to it if bzr didn't move to vcs-noshell?  I realize the
> >> answers are probably obvious to you, but I'm still learning how this
> >> system is set up...
> >
> >Do you understand that all sftp-based services are managed by the
> >'sftp' VM?
> Yes, I think I understand that.  Just to confirm: I can see that cvs,
> hg, git, and svn, are all on the same virtual machine at
> ( -- the 'vcs-noshell' VM -- whereas
> is on (, which is
> presumably 'sftp'.

Yes, it's "sftp".  We don't use the IP (nor the reverse DNS) to
discuss services, we use the VM name instead (some VMs have multiple
IPs, some IPs ports are distributed on different VMs, etc.).

Try to access them using ssh from the Xen Dom0, so
you don't have to "presume" about the architecture :)

> What I don't know is, what is the proper way to migrate bzr services to
> vcs-noshell?  It seems to me like these steps are needed:
>   1) Install bzr and loggerhead on vcs-noshell
>   2) Copy over /srv/bzr/* from sftp.
>   3) Test bzr and loggerhead access with the configuration described in [1]
>   4) Coordinate a flag day with all current bzr users.  Document how
>      bzr+ssh:// access will differ from sftp:// .
>   5) On the flag day, re-sync the latest /srv/bzr data from sftp to
>      vcs-noshell.  This can be done using bzr itself; there is no need
>      to make the old data read-only, since people can just re-push any
>      changes.  The point of the flag day is just to give them a date
>      for which they know to re-check that changes got ported.
>   7) Configure DNS so points to the vcs-noshell machine
>   8) Tell everyone the switchover is done, remind them to check that any
>      changes they pushed after the flag day are in the right place, and
>      to re-push if not.
> Before I ask more questions about any of the steps, is the above roughly
> right?  I could be completely misunderstanding what's needed here...

That sounds OK.

I'm personaly not so sure about asking people to re-push and such, but
up to you.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]