RHSM certguard failure messages are lost in log-level debug
(Copied from BZ)
When a client is denied access by the rhsm-certguard, the log messages describing the reason are lost with log level debug. They should be raised to at least warning, because they tell an administrator, why a client is unable to consume their subscriptions.
Steps to Reproduce:
- Subscribe a host to RH subscription content
- Find a reason for the rhsm content guard to reject the client certificate Either https://bugzilla.redhat.com/show_bug.cgi?id=1977893, or misconfigure it, see below
- On the host run
yum updateand observe that repodata.xml returns 403
- In forman-tail observe that all reasons for ^ are in log level DEBUG
Failed authentication to subscription content should be logged at a higher level, maybe WARNING.
Misconfiguring may be:
curl -vv -k -X PATCH --data-urlencode 'email@example.com' --cert /etc/pki/katello/certs/pulp-client.crt --key /etc/pki/katello/private/pulp-client.key "https://localhost/pulp/api/v3/contentguards/certguard/rhsm/<UUID>/"
BZ comments from ggainey:
I suspect this is a combination of https://github.com/pulp/pulp-certguard/blob/master/pulp_certguard/app/models.py#L42 and https://github.com/pulp/pulp-certguard/blob/master/pulp_certguard/app/models.py#L167-L169.
In the first, we only log "this doesn't even look like a cert" at debug-level.
In the second, we lose information on the specific error encountered and log "something went wrong".
From discussion w/ Matthias:
certguard raises PermissionError, which is a base python error-class. These error-messages appear to not show up in logging. Investigation needed on whether PermissionError is "special" in some way when it comes to being logged by the content-app.
NOTE: we really should not be overloadiung python's file-level PermissionError this way, and instead should have our own cecrtguard-perm-error. May want its own RFE.
See https://github.com/pulp/pulpcore/blob/master/pulpcore/content/handler.py#L296-L303 for where we might want to increase log-level.
Please register to edit this issue