[PATCH conductor 0/2] BZ 785198 - Consistent display of failed images

Matt Wagner matt.wagner at redhat.com
Wed Feb 1 17:43:45 UTC 2012


Hi all,

>From the "exceedingly difficult to track down, but trivial to patch," this patchset fixes what may be an edge case, described in BZ 785198. (Or, at least, it fixes the Conductor side of the problem.) If an image fails while it is building, we may be left with an Image Build and Target Image that are effectively orphaned, but the latest_build attribute is not set on the image because of the failure.

That may be a bug in and of itself (but that is debatable), but the problem in Conductor is that we are entirely inconsistent with how we handle this. We show the image status as "Not built", but the image detail page lists "Provider Images" as if everythign went fine.

I have attached two patches. The first one is perhaps slightly questionable; I include ".enabled" when selecting providers to build for, so that we don't build for disabled providers. This seems correct, and eliminates one source of the root cause of this error. (In my case, I had half-configured a vSphere provider, so building for it blew up spectacularly. The provider was disabled, but we tried to build anyway.) I think this patch enables the correct behavior, but it's possibly up for debate, so it's separate.

The second one is more directly relevant: if we find that we have a target image but @latest_build is not set (and after we have checked for unpushed but built images), we display some text indicating that an error has occurred. I wasn't entirely certain how to word the error, but I tried to err on the side of something concise but useful. I am absolutely open to suggested linguistic changes, but I'm reluctant to display a huge blob of text here.

-- Matt



More information about the aeolus-devel mailing list