Description of problem: I'm trying to build or rebuild my application in https://console.preview.openshift.com/console/project/sspeiche-dotnet/browse/builds/dotnet-sqlite-example I select the button "Start Build" and I get inconsistent behavior. Some times it acts as if I didn't press it (doesn't register the click) but appears to since I can't click anymore. Though there is no indication of a build that has started. I refresh the page and see the build hasn't started. Other times I click "Start Build" and it takes a few seconds to give me a message that is started. Version-Release number of selected component (if applicable): This is preview.openshift.com Version OpenShift Master: v3.2.1.13-5-gddf7d17 Kubernetes Master: v1.2.0-36-g4a3f9c5 How reproducible: Just do repeated builds from Web UI
Reducing as I'm not able to reproduce this very often. It appears the instantiate POST gets stuck in pending and refreshing the browser or waiting will get past this.
Moving to the Rest API, the console is making the request to instantiate the build and as Steve said it sits in Pending. Not happening very often.
Ben, does the instantiate endpoint do anything blocking?
No, if the build object is created, the instantiate endpoint has returned. Jessica, what is sitting in "pending"?
Ok I think I understand the "pending" thing is that the POST call has not returned/finished yet. Steve when you say "the build hasn't started", do you mean the build wasn't *created*?
To better answer Jordan's question: the instantiate api does a bunch of api calls to resolve things related to the buildconfig that we're instantiating a build for, but it should not explicitly block on anything. So the only reason for it to take a long time to respond is if those downstream api calls are slow.
Trying to track it down, though I see these now: api.preview.openshift.com/oapi/v1/namespaces/sspeiche-test3/buildconfigs/njs/instantiate:1 POST https://api.preview.openshift.com/oapi/v1/namespaces/sspeiche-test3/buildconfigs/njs/instantiate 504 (Gateway Timeout)
This honestly is looking like a AWS ELB or HAProxy issue, the pattern I'm starting to see is if I'm away from the browser tab for some time or switch to/from wired/wireless. Ops currently has a pruning job running dealing with long connects connects from ELB and HAproxy, I wonder if the reason of that flakiness is causing this pending and 504 issues.
(In reply to Ben Parees from comment #5) > Ok I think I understand the "pending" thing is that the POST call has not > returned/finished yet. > > Steve when you say "the build hasn't started", do you mean the build wasn't > *created*? It wasn't created (at least I think from latest runs, though I get a 504 error now, maybe I'm just waiting longer to get the actual timeout)
Steve: Is this still an issue? If you are not seeing this anymore (I am not), I would like to move this to ON_QA.
Moving this to ON_QA to retest since the environment has been upgraded to 3.3.x
In Online Dev-preview 3.3, the "Start Build" button always works well in my test: Move this bug to verified. Version OpenShift Master: v3.3.1.3 Kubernetes Master: v1.3.0+52492b4