I always had the impression that stored credentials, be it cached SVN credentials or private keys in people's home dirs, was strongly discouraged. Otherwise I could just have requested an account on vesta and copied the stuff over to people with a private key stored in my home dir on people (see my mail from 2011/02/02). This is why I started this whole thread in the first place...
Uli On 03.03.2011 15:23, Daniel Kulp wrote: > On Thursday 03 March 2011 8:33:21 AM Jukka Zitting wrote: >> Hi, >> >> On Thu, Mar 3, 2011 at 2:30 PM, Ulrich Stärk <u...@apache.org> wrote: >>> The only thing that seems reasonable apart from having an rsync daemon is >>> doing a recursive wget on the generated site every hour or so. This will >>> however fetch all 4000 files and I don't know how much load it puts on >>> Jenkins. Is this OK? >> >> Doesn't sound like a very good approach. >> >> The Apache CMS setup running on Buildbot simply commits the generated >> site to svn from where it's picked up by the public web servers. I >> guess a similar setup could be done also with Maven site builds >> running in Hudson. > > That's kind of what I was thinking. If buildbot has creds stored someplace > to be able to run "svn commit", could one of the Jenkins build slaves > (doesn't > need to be all of them) also have svn installed with stored creds? > > I don't even think it needs to be built into Jenkins itself. If svn is > installed, the build itself (exec plugin or something) could run "svn commit - > m 'Web site update'" or something. > > I think a bunch of projects might be willing to flip to more of a svnpubsub > publish if we could get Jenkins builds to do it. That could reduce some > load > on the rsyncs on p.a.o. > >
