Am 23.06.2012 01:39, schrieb Graham Anderson:
On Monday 18 Jun 2012 11:56:57 Michael Schroeder
What gazillion stat operations? Are you talking
about the scheduler?
Speeding up the scheduler doesn't decrease build time,
The source server +
scheduler operate well until there's an unexpected
problem. I suspect the code doesn't matter, the architecutre is bad.
having too few packages to build is not a problem
we need to solve...
Yet having too many packages seems to be a problem that OBS
About architecture: Last night I had "just an idea"; maybe for the far
future or even impossible...
Anyone remembers SETI@home ?
Could it work, to use the computers of willing community members or
sponsors for OBS build tasks in a way, the SETI@home project uses the
computers of many people and enterprises around the world ?
I have not enough knowledge about the current architecture, to evaluate
this. If it would be necessary to transfer a new virtual machine and ten
thousands of packages for every build task from the server to the
client, that would not work. But if I choose to install at my computer
an appliance for building packages for openSUSE 12.1 x86 or for Fedora
17 x86_64 and get only new/changed source packages for building ?
I think this could work to test updates, i.e. download a snapshot of
factory in a VM and then recompile all sources. But to develop such a
beast would be a huge undertake and then you will find out that only
few people would use that.
But it's basically what happens with samba build farm and perl testers,
To unsubscribe, e-mail: opensuse-factory+unsubscribe(a)opensuse.org
To contact the owner, e-mail: opensuse-factory+owner(a)opensuse.org