Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.

Bug 1834415

Summary:	When following instructions on creating disconnected OLM registry, no packagemanifests are ever created
Product:	OpenShift Container Platform	Reporter:	rvanderp
Component:	OLM	Assignee:	Evan Cordell <ecordell>
OLM sub component:	OperatorHub	QA Contact:	Jian Zhang <jiazha>
Status:	CLOSED DUPLICATE	Docs Contact:
Severity:	high
Priority:	high	CC:	bluddy, sferguso
Version:	4.5
Target Milestone:	---
Target Release:	4.5.0
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2020-05-20 19:21:56 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description rvanderp 2020-05-11 16:43:29 UTC

Description of problem:
After following the instructions at[https://docs.openshift.com/container-platform/4.3/operators/olm-restricted-networks.html], no packagemanifests are ever created.  

Version-Release number of selected component (if applicable):
4.3.8

How reproducible:
Consistently at atleast 2 customers

Steps to Reproduce:
1. Build the catalog image(used --filter-by-os='linux/amd64' otherwise command fails) 
2. Mirror the catalog
3. catalog pod starts and stays started
4. Packagemanifests are never created

Actual results:
packagemanifests are not created

Expected results:
packagemanifests should be created

Additional info:

The catalog pod and catalogsource are both created, yet no package manifests are created.  The only evidence of a problem I can find are these messages in the created catalog pod:

2020-05-06T18:29:17.538499655Z time="2020-05-06T18:29:17Z" level=warning msg="couldn't migrate db" database=/bundles.db error="attempt to write a readonly database" port=50051
2020-05-06T18:29:17.53955741Z time="2020-05-06T18:29:17Z" level=info msg="serving registry" database=/bundles.db port=50051

I also extracted bundles.db from the catalog image and queried 'select * from operator bundle' which seemed to return information on each bundle.

Comment 1 Ben Luddy 2020-05-13 19:22:00 UTC

Hi Richard,

Can you share logs from the packageserver deployment? The catalog pod logs look okay to me, and it sounds like the database is not empty, so it's possible that the package server (which is responsible for the package manifest resource) may be having trouble syncing with the catalog pod.

You might also try rolling packageserver -- after collecting logs -- on the off chance that it solves the immediate issue.

Comment 3 Evan Cordell 2020-05-18 14:20:42 UTC

This may indicate a version mismatch between oc / the base image / the target ocp cluster.

Can you please share the version of oc, the base image used (--from in oc adm catalog build) and confirm the version of the target cluster?

If you find that they don't match, please try oc adm catalog build again, but with oc and baseimage for the target ocp cluster.

Comment 4 rvanderp 2020-05-18 21:17:44 UTC

We are following up on this now.  Should hopefully have an update in the morning.  Thanks!

Comment 5 Evan Cordell 2020-05-20 19:21:56 UTC

Because of the log "couldn't migrate db", I now suspect this is a dupe of https://bugzilla.redhat.com/show_bug.cgi?id=1827612

I will mark it as such for simpler tracking, but if we discover there is an issue in addition the the linked bug, please re-open with more details and we will continue to investigate.

*** This bug has been marked as a duplicate of bug 1827612 ***