sap_hana_preconfigure: Support loading vars which are valid for RHEL minor release and all later ones

[berndfinger]

Ansible Role

sap_hana_preconfigure

OS Family

RHEL

Feature/Enhancement request

We want to be able to define role vars which are valid for a certain RHEL minor release and all later ones, without having to define one vars file per each RHEL minor release. This has been implemented as part of #1083 but is causing the problem initially reported in #1093, see also #1094.
If possible, avoid loading role vars files via a find task running on localhost, see #1094 (comment).

The order of variable loading should be as follows (example for SLES 15.6 or SLES_SAP 15.6 and RHEL 9.4):

1. Suse.yml / RedHat.yml
2. SLES_15.yml / RedHat_9.yml
3. (SLES_SAP only) SLES_SAP_15.yml
4. (RHEL only) RedHat_9.1.plus.yml
5. (RHEL only) RedHat_9.2.plus.yml
6. (RHEL only) RedHat_9.3.plus.yml
7. (RHEL only) RedHat_9.4.plus.yml
8. SLES_15.6.yml / RedHat_9.4.yml
9. (SLES_SAP only) SLES_SAP_15.6.yml

Variables defined in later vars files will overwrite those defined before.

Note: Redefining a variable for a specific RHEL minor release (e.g. RedHat_9.4.yml) which is already defined in a plus file for the same RHEL minor release (e.g. RedHat_9.4.plus.yml) should be avoided because otherwise, users might have to investigate the variable loading order.

Community participation

Happy to implement this feature

Optional - Attempted solutions or manual steps

No response

[marcelmamula]

We should keep existing var file selection and RHEL specific (minor versions and _plus versions) can be handled as separate loading inside of RHEL specific task as currently with find code.

[berndfinger]

We should keep existing var file selection and RHEL specific (minor versions and _plus versions) can be handled as separate loading inside of RHEL specific task as currently with find code.

For RHEL, we want to load the plus files after the major release and before the minor release. This is not possible if we leave the existing include_vars task as it is.

I see the following options:

1. We use the existing include_vars task only for Suse and implement the new logic in other tasks which are only valid for RedHat.
2. We continue using the existing include_vars task for both Suse and RedHat but only for major versions and move the minor release specific var file loading to another include_vars task also for both Suse and RedHat. We then can add a third include_vars task for RHEL _plus in between.

[marcelmamula]

I said leave it as it is and replicate same inside of RHEL specific task that runs afterwards, it will load and replace vars when found, while leaving existing behavior unchanged.

This is why we designed it to be additive, where base OS Family vars are baseline, and Major/Minor vars are adding on top of it.

[berndfinger]

I said leave it as it is and replicate same inside of RHEL specific task that runs afterwards, it will load and replace vars when found, while leaving existing behavior unchanged.

If we leave it as it is, and replicate the same inside of a RHEL specific task that runs afterwards, the following can happen (example for RHEL 9.2, with files RedHat_9.yml, RedHat_9.2.yml, and RedHat_9.2.plus.yml):

1. In the first include_vars task, we load the variables which are defined in RedHat_9.yml and if the same variable is defined in RedHat_9.2.yml, we use that one.
2. We then load the variables which are defined in RedHat_9.yml again, then replace one of those by those which defined in RedHat_9.2.plus.yml, then replace it by the one defined in RedHat_9.2.yml.

That does not make much sense to me.

[berndfinger]

Fixed in #1097.