11:16 <ElectricJozin> I don't get how you are suppose to specify the session bus address in the Dbus C api. I think thats why this code doesnt work: https://pastebin.com/7xc4rYBU. I am getting this output in dbus-monitor: https://pastebin.com/qbmjJPjz The bus id printed by dbus_bus_get_id is different that of the session bus.

11:17 <ElectricJozin> By "not work" I mean it doesnt print a notification as oppose to calling the method from d-feet

11:20 <emersion> this channel is about the freedesktop infrastructure, ie. gitlab, mailing lists, etc

14:34 <bentiss> daniels: quick FYI, I'm going to set the pull policy on the gitlab runners to if-not-present, the default is always, and maybe we don't need that

14:34 <daniels> bentiss: yeah that sounds like a great idea

14:41 <bentiss> daniels: alright, it's done, We'll see if that has an effect on the GCS pulls

15:18 <alatiera> I don't think it will have an effect

15:18 <alatiera> its mostly useful if you have set tags that you push new revisions to, but we mostly have a tag per build with the template

15:19 <alatiera> so for example foobar:latest or foo:1.22 images

15:20 <alatiera> even with always the image still gets cached and stays there, its mostly a secondary check for the object sha matching the one in published in the registry

15:53 <bentiss> alatiera: seeing https://harbor.freedesktop.org/harbor/projects/7/repositories and ordering by pulls, you'll see that mesa/mesa/debian/x86_build has been pulled 60322 times for instance, the runners are *constantly* pulling the whole images, as we see on the GCP logs

15:54 <bentiss> an other way of seeing that: the cache for the past 15 days is 237.24 GiB, when the total pull over 7 days is more than 1 TB. We have arguably more than one runner, but still, we would have expected a little bit more caching

15:58 <alatiera> yea those numbers look weird

15:59 <alatiera> I wonder if we can know which tags is the one pulled most and when

16:24 <bentiss> the numbers of pulls on harbor doesn't necessary reflect the actual pulls of blobs. But on grafana we can get the culprit quite easily.