[opensuse-factory] build power
Hi all, In a couple of weeks we expect a new server rack, sponsored by SUSE, to arrive. It will bring enough build power to solve the long waiting many of you are experiencing. See http://news.opensuse.org/?p=15739 for details. In the mean time, cleaning up unused packages and projects in your home always helps us save some cycles! Our binary delivery server is also getting full and that, too, will need a solution. Cleaning up only gets us so far. Help finding a new server is welcome, Adrian can probably tell you what we need if you ask. Remember, this service depends on support from sponsors and we can use more help. Cheers, Jos
On Mon, Apr 8, 2013 at 8:05 AM, Jos Poortvliet
Hi all,
In a couple of weeks we expect a new server rack, sponsored by SUSE, to arrive. It will bring enough build power to solve the long waiting many of you are experiencing. See http://news.opensuse.org/?p=15739 for details.
In the mean time, cleaning up unused packages and projects in your home always helps us save some cycles!
I don't want to sound ungrateful, on the contrary, but aren't those 5.4k rpm drives a bad choice? A great many times builds are limited by IOPS, in my limited experience. That means, in those many cases, those low-rpm drives will be a pretty narrow bottleneck. Has OBS worker load been analyzed to get to the conclusion that 5.4k rpm drives are a good choice? I'd be interested on seeing that analysis, since at some place I used to work we had a mini-farm that could benefit from some of that insight (ie - I'd like to forwarding that to my former sysadmin ;^) ) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
Hi
Claudio Freire
I don't want to sound ungrateful, on the contrary, but aren't those 5.4k rpm drives a bad choice?
We took the sponsored system just "as it is" - if we got the possibility, we probably would even not use any local hard disk and have everything in the (bigger) RAM of the machine. Now we will see how many workers can be started in parallel - and afterwards divide the available RAM and see if there is enough left to start the workers in a tmpfs image. If I had the choice, I would * use tmpfs for building * use SSDs for the OBS cache as we boot the workers via PXE, there needs no pre-installed OS on the hard disks. And there is nothing that can replace RAM - beside more RAM ;-) I hope that answers your questions. In general you are right, but we take what we can get and will try to get the best out of it... ;-) With kind regards, Lars -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Mon, Apr 8, 2013 at 4:12 PM, Lars Vogdt
If I had the choice, I would * use tmpfs for building * use SSDs for the OBS cache
as we boot the workers via PXE, there needs no pre-installed OS on the hard disks. And there is nothing that can replace RAM - beside more RAM ;-)
I hope that answers your questions. In general you are right, but we take what we can get and will try to get the best out of it... ;-)
It does, thanks. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Monday 2013-04-08 13:05, Jos Poortvliet wrote:
Our binary delivery server is also getting full and that, too, will need a solution. Cleaning up only gets us so far. Help finding a new server is welcome, Adrian can probably tell you what we need if you ask.
With respect to multi-threading, the source code could use improvements in some places. I have seen bs_srcserver using multiple processes (at least something!) at times, but that does not seem to be the case for bs_repserver or bs_sched. Especially the scheduler would benefit from it in cases where #cores >> #architectures to schedule for. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, Apr 09, 2013 at 12:25:15AM +0200, Jan Engelhardt wrote:
On Monday 2013-04-08 13:05, Jos Poortvliet wrote:
Our binary delivery server is also getting full and that, too, will need a solution. Cleaning up only gets us so far. Help finding a new server is welcome, Adrian can probably tell you what we need if you ask.
With respect to multi-threading, the source code could use improvements in some places. I have seen bs_srcserver using multiple processes (at least something!) at times,
You're kidding, right? The source server is a web server, it'll spawn a new process for every request.
but that does not seem to be the case for bs_repserver or bs_sched. Especially the scheduler would benefit from it in cases where #cores >> #architectures to schedule for.
That depends on if it's IO bound or CPU bound. If it's CPU bound, another option is to improve the code. We already do some of the heavy work in an XS module, be might want to move some other stuff there as well. In most of the cases it seems to be IO bound, though. Note also that the current version of the scheduler supports async src server requests, we might want to try that for OBS. Cheers, Michael. -- Michael Schroeder mls@suse.de SUSE LINUX Products GmbH, GF Jeff Hawn, HRB 16746 AG Nuernberg main(_){while(_=~getchar())putchar(~_-1/(~(_|32)/13*2-11)*13);} -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
participants (5)
-
Claudio Freire
-
Jan Engelhardt
-
Jos Poortvliet
-
Lars Vogdt
-
Michael Schroeder