Chapter 4. Known issues

The current version of cosign expects only one single Timestamp Authority (TSA) certificate chain. When rotating the TSA certificate chain, you give the whole TSA certificate chain to The Update Framework (TUF) as an individual target. During the rotation process, setting the new TSA certificate chain as the new TUF target, and expiring the old TSA certificate chain gives the following error message.

main.go:74: error during command execution: unable to load TSA certificates: TSA certificate chain must contain exactly one leaf certificate

Currently, there is no workaround for this issue.

For information about rotating the TSA signer key and certificate chain see our procedure for Red Hat OpenShift Container Platform, or Red Hat Enterprise Linux.

When specifying a persistent volume claim (PVC) name in The Update Framework (TUF) resource causes the RHTAS operator to fail the initialization of the TUF repository. For example:

spec:
...
  tuf:
    ...
      pvc:
        name: tuf-pvc-example-name
...

spec:
...
  tuf:
    ...
      pvc:
        name: tuf-pvc-example-name
...

To workaround this issue, do not specify a PVC name in the TUF resource. This allows the RHTAS operator to automatically create the PVC, names it tuf, and properly initializes the TUF repository.

After upgrading the RHTAS operator to the latest version (1.0.1), the existing Rekor data is not found when searching by email address. The backfill-redis CronJob, which ensures that Rekor Search UI can query the transparency log only runs once per day, at midnight. To workaround this issue, you can trigger the backfill-redis job manually, instead of waiting until midnight.

To trigger the backfill-redis job from the command-line interface, run the following command:

oc create job --from=cronjob/backfill-redis backfill-redis -n trusted-artifact-signer

$ oc create job --from=cronjob/backfill-redis backfill-redis -n trusted-artifact-signer

Doing this adds the missing data back to the Rekor Search UI.

We found a potential issue with the RHTAS Operator logic that can cause an unexpected state when redeploying. This inconsistent state can happen if removing configurations from RHTAS resources and the Operator tries to redeploy those resources. To workaround this potential issue, you can delete the specific resource, and then re-create that resource by using the previous instance’s configuration, such as keys, and persistent volumes. The RHTAS resources are: Securesign, Fulcio, The Update Framework (TUF), Rekor, Certificate Transparency (CT) log, or Trillian.

For example, to delete the Securesign resource:

oc delete Securesign securesing-sample

$ oc delete Securesign securesing-sample

For example, to re-create the Securesign resource from a configuration file:

oc create -f ./securesign-sample.yaml

$ oc create -f ./securesign-sample.yaml

When restoring the RHTAS signer data from a backup to a new OpenShift cluster, the component status links do not update as expected. Currently, you have to manually delete the securesign-sample-trillian-db-tls resource, and manually update the component status links. The RHTAS operator will automatically re-create an updated securesign-sample-trillian-db-tls resource, after it has been removed.

After the backup procedure starts, and the secrets restored, delete the securesign-sample-trillian-db-tls resource:

oc delete secret securesign-sample-trillian-db-tls

$ oc delete secret securesign-sample-trillian-db-tls

Once all the pods start, then update the status files for Securesign, and TimestampAuthority:

oc edit --subresource=status Securesign securesign-sample
oc edit --subresource=status TimestampAuthority securesign-sample

$ oc edit --subresource=status Securesign securesign-sample
$ oc edit --subresource=status TimestampAuthority securesign-sample