[opensuse-buildservice] armv7 workers
Hi, Seems the armv7 waiting queue is quite long now. And there is only a few native armv7 workers, at the same time it seems there is no qemu-based-armv7 workers. At the same time, seems that all of armv6 workers are qemu. The question is what is the policy not to run armv7 on x86_64 workers? -- To unsubscribe, e-mail: opensuse-buildservice+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-buildservice+owner@opensuse.org
Hi Matwey,
The question is what is the policy not to run armv7 on x86_64 workers?
Its not a policy decision per se, its a "lesser of two evils" decision. There are quite often regressions / build failures happening only due to emulation bugs, and it is easier to avoid all that hassle by building natively. Whenever we're hit by such an issue, we often also require somebody with admin privileges to repair the damage far enough so that we can actually continue building. With native builds, we have more control and can inject workaround packages ourselves. Also, surprisingly building natively is often faster, at least for packages that are not CPU bound. Now coming back to your original question: The main reason for the long build queue was that the main build hardware has crashed including the remote mgmt interface and it took us to find somebody with physical access to the machine room in order to power cycle it. now that this has been restored after a few weeks of waiting, the build queue is down to the minimum. There is a certain number (~200) of build jobs that will never be built as they have constraints that are not fulfillable, so you just need to discount that from the actual numbers. We even have idle workers right now. Greetings, Dirk -- To unsubscribe, e-mail: opensuse-buildservice+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-buildservice+owner@opensuse.org
Ok, it is clear now. have you seen http://www.dell.com/learn/us/en/555/campaigns/project-copper ? 03.01.2014 17:18, Dirk Müller пишет:
Hi Matwey,
The question is what is the policy not to run armv7 on x86_64 workers?
Its not a policy decision per se, its a "lesser of two evils" decision. There are quite often regressions / build failures happening only due to emulation bugs, and it is easier to avoid all that hassle by building natively. Whenever we're hit by such an issue, we often also require somebody with admin privileges to repair the damage far enough so that we can actually continue building. With native builds, we have more control and can inject workaround packages ourselves.
Also, surprisingly building natively is often faster, at least for packages that are not CPU bound.
Now coming back to your original question: The main reason for the long build queue was that the main build hardware has crashed including the remote mgmt interface and it took us to find somebody with physical access to the machine room in order to power cycle it. now that this has been restored after a few weeks of waiting, the build queue is down to the minimum. There is a certain number (~200) of build jobs that will never be built as they have constraints that are not fulfillable, so you just need to discount that from the actual numbers.
We even have idle workers right now.
Greetings, Dirk
-- To unsubscribe, e-mail: opensuse-buildservice+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-buildservice+owner@opensuse.org
Hi Matwey,
it is clear now. have you seen http://www.dell.com/learn/us/en/555/campaigns/project-copper ?
Yes, we saw it, but it is Armada XP based, which we threw out roughly a year ago (due to being too slow). Our current build hardware is way faster than that machine already. Greetings, Dirk -- To unsubscribe, e-mail: opensuse-buildservice+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-buildservice+owner@opensuse.org
participants (2)
-
Dirk Müller -
Matwey V. Kornilov