- Performed the following on the ‘Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server’ and ‘Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server’ repos:
    - Verify Content Checksum
    - Advanced Sync -> Complete Sync

  Outcome:

  - [satellite] Sync Summary for Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server

SYNC SUMMARY
The synchronization of "Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server" has completed. Below is a summary of new errata. A large number of errata were synced for this repository, so only the first 100 are shown.

New Errata:
200  SECURITY
222  BUGFIX
36   ENHANCEMENT


  - [satellite] Sync Summary for Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server

The synchronization of "Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server" has completed. Below is a summary of new errata. A large number of errata were synced for this repository, so only the first 100 are shown.

New Errata:
206   SECURITY
358   BUGFIX
67    ENHANCEMENT


Whereas the sync status page in Satellite's web UI shows:

Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server	22 minutes ago	9 minutes	No content added.	Syncing Complete.
Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server			22 minutes ago	19 minutes	No content added.	Syncing Complete.

So, even with Advanced Sync -> Complete Sync of the 2 repos, the errata count in sync status page in satellite 6.10 web UI and e-mail notifications for the manually initiated completed Advanced Sync -> Complete Syncs does not align.

On 24.1.2022:
-------------

- [satellite] Sync Summary for Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server

SYNC SUMMARY
The synchronization of "Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server" has completed. Below is a summary of new errata.

New Errata:
1  SECURITY
0  BUGFIX
0  ENHANCEMENT


- [satellite] Sync Summary for Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server

SYNC SUMMARY
The synchronization of "Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server" has completed. Below is a summary of new errata.

New Errata:
1  SECURITY
0  BUGFIX
0  ENHANCEMENT

So, the following actions fixed the issue temporarily:
  -Performed the following on the ‘Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server’ and ‘Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server’ repos:
    - Verify Content Checksum
    - Advanced Sync -> Complete Sync

But again, on next run of the repo sync plans, the issue re-appeared:

On 25.1.2022:
-------------

- [satellite] Sync Summary for Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server

SYNC SUMMARY
The synchronization of "Red Hat Enterprise Linux 7 Server RPMs x86_64 7Server" has completed. Below is a summary of new errata. A large number of errata were synced for this repository, so only the first 100 are shown.

New Errata:
207  SECURITY
358  BUGFIX
67  ENHANCEMENT


- [satellite] Sync Summary for Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server

SYNC SUMMARY
The synchronization of "Red Hat Enterprise Linux 7 Server - Optional RPMs x86_64 7Server" has completed. Below is a summary of new errata. A large number of errata were synced for this repository, so only the first 100 are shown.

New Errata:
201  SECURITY
222  BUGFIX
36   ENHANCEMENT

Hi, 

It looks like the errata mailer is just sending the list of all errata in the repository, which is what i thought it did, and was surprised that its telling the user that its 'New Errata'.  That is incorrect.

Ideally we change this to actually return the newly synced errata, at a bare minimum, we should at least make the email text correct.

(In reply to Justin Sherrill from comment #4)
> Hi, 
> 
> It looks like the errata mailer is just sending the list of all errata in
> the repository, which is what i thought it did, and was surprised that its
> telling the user that its 'New Errata'.  That is incorrect.
> 
> Ideally we change this to actually return the newly synced errata, at a bare
> minimum, we should at least make the email text correct.

Katello is behaving properly here. It queries pulp prior and after the sync and what is new or different(!), it marks as a New Errata. And imho the mail notification feature *should* be sending *just new/modified* errata, not all.

The "bad guy" here is pulp that - when detecting there is an errata to newly sync - it creates records for *all* errata from the repo and update the repository to use the new records. That confuses katello.

See example:

katello_repository_errata table :
  id  | erratum_id | repository_id |     created_at      |     updated_at      |                            erratum_pulp3_href
 2914 |          1 |             2 | 2022-06-29 07:28:08 | 2022-06-29 07:28:08 | /pulp/api/v3/content/rpm/advisories/b84ca7e0-1669-435d-90fe-db95df1fd7df/

changed to:

 2914 |          1 |             2 | 2022-06-29 11:49:13 | 2022-06-29 11:49:13 | /pulp/api/v3/content/rpm/advisories/62eb7f9e-e5d0-4956-ad83-611fd8639b13/

both advisories URLs contain the *same* errata, these are duplicate records. The new errata was *created* during the new sync (that synced/updated just one different errata).


And from pulp side: 
1) having a repo with 1 errata synced
2) update the remote repo to have some more packages and *one* another errata
3) check # errata in pulp:

certs="--cacert /etc/pki/katello/certs/katello-server-ca.crt --cert /etc/pki/katello/certs/pulp-client.crt --key /etc/pki/katello/private/pulp-client.key"
hname=$(hostname -f)
curl -s $certs https://${hname}/pulp/api/v3/content/rpm/advisories/ | json_reformat | head
{
    "count": 4940,
..

4) sync the repo to fetch the new RPMs and some errata (one new here)
5) check # errata in pulp:

curl -s $certs https://${hname}/pulp/api/v3/content/rpm/advisories/ | json_reformat | head
{
    "count": 4942,
..

Just *one* new errata appeared in the repo, but *two* new entries there.

6) repeat it once more by adding *one* another new errata to sync:

curl -s $certs https://${hname}/pulp/api/v3/content/rpm/advisories/ | json_reformat | head
{
    "count": 4945,
..

and *three* errata appears! All three from the repo.



Satellite's reproducer:
1) Put to /var/www/html/pub/ the three directories tools_repodata_[1-3] I will attach. These are repodata of RHEL7 Sat6.9 tools gradually adding one errata at a time.
2) create a product and a repo:
hammer product create --name "tools_test_product" --organization-id 1
hammer repository create --organization-id 1 --product tools_test_product --name tools_repo_1 --content-type yum --url http://localhost/pub/tools_repodata --download-policy on_demand

3) point tools_repodata to the first "version" of the repodata:
i=1; rm -f /var/www/html/pub/tools_repodata; ln -s /var/www/html/pub/tools_repodata_${i} /var/www/html/pub/tools_repodata; chown -R apache:apache /var/www/html/pub/tools*

4) sync the repo:
hammer repository sync --organization-id 1 --product tools_test_product --name tools_repo_1

5) Check the number of advisories in pulp (and for katello, check records in katello_repository_errata table in foreman DB)

6) point tools_repodata to the second "version" of the repodata:
i=1; rm -f /var/www/html/pub/tools_repodata; ln -s /var/www/html/pub/tools_repodata_${i} /var/www/html/pub/tools_repodata; chown -R apache:apache /var/www/html/pub/tools*

5) sync the repo again and check pulp advisories count - increased by *two* (and the repo<->advisories mapping in pulp point to the *new* advisory records - what also katello_repository_errata confirms)

6) repeat the test with third "version" of the repo


Needinfo to David to explain the pulp behaviour.

Created attachment 1893438 [details]
tools repodata for tiny reproducer

Three versions of the repodata for a reproducer.

I forgot to mention that when a repo sync finds nothing to sync (no errata or no new content at all? I dont know this), then no new advisory is created (which is expected).

Simply a new content (or advisory in particular)? detected during a repo sync is the mandatory condition of the reproducer / bug.

(not 100% sure if this is correct, just indirect evidences for this)

For newly synced repositories on Sat6.10, all the errata are added to the mail notification / all errata are duplicated in pulp advisories.

For content originally migrated from pulp-2, the errata migrated from pulp-2 are *not* included in the mail notification / they are not duplicated among pulp advisories. (this is just indirectly observed).

Closing this as a duplicate of https://bugzilla.redhat.com/show_bug.cgi?id=2044314 that *should* fix the errata mails (though my attempt to backport to 6.10 failed: I still saw all mails there).

The reason why pulp creates new advisory record is that the same errata has subtle differences in the three repodata updateinfo details - so pulp calculates a different digest for the errata. The old errata is removed during orphans removal, so no "advisories leak" happens.

*** This bug has been marked as a duplicate of bug 2044314 ***