On Freitag 28 November 2008 16:39:12 Andrew Beekhof wrote:
On Nov 28, 2008, at 3:19 PM, Adrian Schröter wrote:
On Freitag 28 November 2008 14:32:55 Andrew Beekhof wrote:
On Nov 28, 2008, at 1:43 PM, Adrian Schröter wrote:
On Freitag 28 November 2008 13:27:43 Andrew Beekhof wrote:
All jobs in https://build.opensuse.org/project/monitor?project=home%3Abeekhof were in the "finished' state for nearly an hour.
The x86_64 are okay, because the scheduler was reloading to take the new code we need. (this takes about two hours with the size of our service). We do this for this reason per architecture, so only one is hanging for a while.
The i586 scheduler has processed it meanwhile, I saw that it had more than 2000 medium events some hours ago (means that it has to recalculate 2000 project repositories due to changed packages). If you can provide a patch which makes this faster, I am happy to apply. (small sidenote, a compiler which compiles the entire distribution instantly would be also appricated :).
You're missing the point. This is not about scheduler reloads taking too long.
They're necessary, we get that... they take a while, we get that too... just tell us when they (and other maintenance events) are happening.
You can see this always on the monitor page.
No. You can't.
The last i586 scheduler restart was 4 days ago and the dispatcher/ publisher was also over a day. So nothing on that page would indicate why i586 packages were not moving from the finished to the success state around lunch time today.
Well it was restarted some minutes ago, but at the time of your writing you was able to see that the x86_64 scheduler got restarted. The reason why the i586 scheduler was slow are just > 600 source package changes (submitted within 10 minutes) and more than 1600 package build events. Yes, we look permanently how to improve the speed. All what I wanted to say that we do not do stupid things by intention and it is not obviously easy to fix this within seconds.
We have often enough multiple thousands packages to be build, waiting for them will quite often not work (and this is independend of any kind of maintenance).
The status page is the first thing I look at when something seems to be taking a while (when I checked today the queue lengths were single digits) and if I see a high load the immediate reaction is: "Oh, its just really busy. I'll do something else then".
Which again illustrates my point - if we know what's going on, we'll be more understanding.
Look again, you see that the i586 scheduler was restarted some minutes ago. Yes, it would be nice to see also the open events to get processed.
Is it really so much trouble to add a one-liner to the status messages?
The message can (and will) be forgotten in daily work (it is not really the case that we only work on one thing in parallel). A automatic report, like it can be found there with the starting time of the scheduler can't :) -- Adrian Schroeter SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg) email: adrian@suse.de -- To unsubscribe, e-mail: opensuse-buildservice+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-buildservice+help@opensuse.org