[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Repository failover
Mark D. Baushke
Re: Repository failover
Sat, 15 Mar 2003 23:46:04 -0800
Mahantesh M Vantmuri <address@hidden> writes:
> Mark wrote:
> >Haven't looking into this much, but if CVSup can send based on date,
> >maybe it can be used to assist failover. Any thoughts from those that
> >have used/are using CVSup?
> We are using cvsup as a repository syncing tool. The cvsup client,
> running on machine having local repositiry, gets the changes that were
> made on the remote repository every one hour. As others mentioned
> earlier in this thread, the local repository becomes read-only which
> means we can use it for only updates but not for commits. If we have
> to commit we do need to change our CVSROOT to the remote repository
> through some tcl script.
Or, you can just specify the remote repository with the -d argument to
In addition, if you use the CVS_LOCAL_BRANCH_NUM patch that has been
committed to the cvshome main trunk (already available for various of
the *BSD versions of cvs), you can create a local development area
branch that the CVSup will not touch, this allows you to develop a
feature 'locally' and then when you have tested it and are ready for it
to be part of the main repository, merge it into the main trunk and
commit using the -d trick.
> So in the similar way we can have failover repository by using CVSUP.
> Have two repositories on two different machines one being read-only,
> and the other write also(but end users or users can completely ignore
> the existance of the backup repository). Have one cron job to pull
> changes done on main repository. In case something happens to main
> repository, you can immediately, stop CVSUP activities on backup cvs
> server, and change permissions on modules on backup cvs server and
> put it on to production as mail repository(Do remember that one needs
> to again setup CVSUP server and similar backup server for this server
> which just now came as production server for CVS)
Hmmm... It sounds tricky to know when to stop your CVSup activities
unless you just lose connectivity to the main repository.
> I hav'nt come acoss clustering kind of CVS servers which will take
> care of repositories as soon as one of the servers gets corrupted. Do
> have a look as DCVS also.
I think you would be better off to keep your main repository as the
primary and have the secondary 'backup' cvs servers available for some
local CVS_LOCAL_BRANCH_NUM work until the main repository is back