Bug 2227902 - Satellite Capsule clamps reply to a max of 100 tags when listing the tags (break RHOSP)
Summary: Satellite Capsule clamps reply to a max of 100 tags when listing the tags (br...
Keywords:
Status: VERIFIED
Alias: None
Product: Red Hat Satellite
Classification: Red Hat
Component: Container Management - Content
Version: 6.12.4
Hardware: x86_64
OS: Linux
urgent
high
Target Milestone: 6.12.5
Assignee: Ian Ballou
QA Contact: Vladimír Sedmík
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2023-07-31 19:27 UTC by Odilon Sousa
Modified: 2023-08-11 21:18 UTC (History)
5 users (show)

Fixed In Version: rubygem-smart_proxy_container_gateway-1.0.8-1.el8sat
Doc Type: If docs needed, set a value
Doc Text:
Clone Of: 2221407
Environment:
Last Closed:
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Foreman Issue Tracker 36616 0 Urgent Closed [smart_proxy_container_gateway] Smart proxy clamps reply to a max of 100 tags when listing the tags 2023-08-02 16:01:29 UTC
Red Hat Issue Tracker SAT-18855 0 None None None 2023-07-31 19:27:34 UTC
Red Hat Issue Tracker SAT-19287 0 None None None 2023-07-31 19:30:16 UTC

Description Odilon Sousa 2023-07-31 19:27:21 UTC
+++ This bug was initially created as a clone of Bug #2221407 +++

Description of problem:

We're using Satellite 6.12.4.1 (Satellite + Capsule) to serve RHOSP content (rpms and containers).

I've noticed that the capsule -clamps- the list of tags to 100 and the sonsequence is that this breaks 'openstack undercloud install' and the automation around it.

Version-Release number of selected component (if applicable):
6.12.4.1 on RHEL8.8

How reproducible:
100%

Steps to Reproduce:
1. Add RHOSP16 container repos to a satellite CV, promote and sync to the capsule.
2. List the container tags from the Satellite + Capsule.
3. All of the replies from the capsule are clamped to '100'.

* Satellite:
# skopeo list-tags docker://w0617ialabrhsat01.iaas.tdlab.ca:443/test-openstack-test_osp16-nfvi-evo-osp16-nfvi-evo-osp16-openstack-nova-api|grep '16\.[0-9]'|wc -l
102
# skopeo list-tags docker://w0617ialabrhsat01.iaas.tdlab.ca:443/test-openstack-test_osp16-nfvi-evo-osp16-nfvi-evo-osp16-openstack-nova-compute-ironic|grep '16\.[0-9]'|wc -l
112


* Capsule:
# skopeo list-tags docker://satellite-capsule01.maas:443/test-openstack-test_osp16-nfvi-evo-osp16-nfvi-evo-osp16-openstack-nova-api|grep '16\.[0-9]'|wc -l
100
# skopeo list-tags docker://satellite-capsule01.maas:443/test-openstack-test_osp16-nfvi-evo-osp16-nfvi-evo-osp16-openstack-nova-compute-ironic|grep '16\.[0-9]'|wc -l
100

--- Additional comment from  on 2023-07-08T23:01:59Z 

The issue here is on Capsules only. On Satellite, the results seem to be retrieved by Foreman from it's own DB. On a Capsule that doesn't exist so foreman-proxy asks Pulp to return the list. Once there the request for tags/list defines this `pagination_class = ContainerTagListPagination`, which just loops back to `api_settings.PAGE_SIZE:`, and we can find this in;
~~~
[root@capsule1 ~]# grep PAGE_SIZE /usr/lib/python3.9/site-packages/pulpcore/app/settings.py
    "PAGE_SIZE": 100,
~~~

You can see there Pulp defaults to returning 100 lines per page, and we need to increase that to get beyond this problem, at least temporarily.

--- Additional comment from  on 2023-07-10T13:13:41Z 

The fix can likely be to introduce paging here: https://github.com/Katello/smart_proxy_container_gateway/blob/main/lib/smart_proxy_container_gateway/container_gateway_main.rb#L46

--- Additional comment from  on 2023-07-26T21:42:51Z 

Created redmine issue https://projects.theforeman.org/issues/36616 from this bug

--- Additional comment from  on 2023-07-27T00:03:15Z 

Upstream bug assigned to iballou

--- Additional comment from  on 2023-07-27T00:03:17Z 

Upstream bug assigned to iballou

Comment 1 Bryan Kearney 2023-08-02 16:03:31 UTC
Moving this bug to POST for triage into Satellite since the upstream issue https://projects.theforeman.org/issues/36616 has been resolved.

Comment 2 Vladimír Sedmík 2023-08-08 08:01:47 UTC
Verified in 6.12.5 snap 2 (with rubygem-smart_proxy_container_gateway-1.0.8-1.el8sat.noarch)

Steps:
1) Sync some big container repo to Satellite. (I used jetstack/cert-manager-controller from quay.io which had 821 tags)
2) Register an external Capsule to the Satellite, add to Library LCE and sync it.
3) Using skopeo, list the tags on both instances and compare:
[user@localhost ~]$ skopeo list-tags --tls-verify=false --username=admin --password=toviszejo docker://satellite.redhat.com/bz-2227902-jetstack-jetstack_cert-manager-controller | grep '"' | wc -l
823
[user@localhost ~]$ skopeo list-tags --tls-verify=false --username=admin --password=toviszejo docker://capsule.redhat.com/bz-2227902-jetstack-jetstack_cert-manager-controller | grep '"' | wc -l
823
[user@localhost ~]$ diff <(skopeo list-tags --tls-verify=false --username=admin --password=toviszejo docker://satellite.redhat.com/bz-2227902-jetstack-jetstack_cert-manager-controller | sort) \
> <(skopeo list-tags --tls-verify=false --username=admin --password=toviszejo docker://capsule.redhat.com/bz-2227902-jetstack-jetstack_cert-manager-controller | sort)
608c608
<     "Repository": "satellite.redhat.com/bz-2227902-jetstack-jetstack_cert-manager-controller",
---
>     "Repository": "capsule.redhat.com/bz-2227902-jetstack-jetstack_cert-manager-controller",

Expected and actual results:
Both instances provided the same tags, the count was higher than 100 and matched the expected tags count.

Note:
The `wc -l` gave two more since we got these two extra lines, which is expected:
   "Repository": "capsule.redhat.com/bz-2227902-jetstack-jetstack_cert-manager-controller",
   "Tags": [


Note You need to log in before you can comment on or make changes to this bug.