Currently Ceph subs doesn't include ha repo, and we don't expect any packages like pacemaker from ha repo used in Ceph storage nodes, IIUC.

Note that we have the same issue in our installation doc, too.
 https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.1/html/director_installation_and_usage/planning-your-overcloud#overcloud-repositories

It seems that we have one issue with upgrade when ha repo is disabled. See bz1899404 to find the detail.

Then should we add ha repo to Ceph Subs ?

Hi Takashi-san,

So this is largely dependent on the type of subscription used and how the Ceph Storage nodes were deployed. There are two scenarios:

1. The customer is using separate OpenStack and Ceph subscriptions (applied to the specific nodes) and the Ceph Storage nodes were deployed using the overcloud-minimal image. In this case, the HA repo wouldn't be used because the Ceph Storage nodes are most likely using native linux networking management.

2. The customer has a combined OpenStack and Ceph subscription (Service Provider SKUs and HCI SKUs) and the Ceph Storage nodes were deployed using the overcloud-full image. In this case, you would need to include the HA repo because the Ceph Storage nodes are most likely using openvswitch to manage networking and the HA repo is needed to upgrade openvswitch.

It's probably not a good idea to remove the repo but maybe a better idea to clarify the usage. What do you think?

Hi, Dan,

As I mentioned above openvswitch is included in fast-datapath-for-rhel-8-x86_64-rpms,
not in rhel-8-for-x86_64-highavailability-rpms .
I'm not aware of any usecase where standalone CephStorage nodes require packages
from rhel-8-for-x86_64-highavailability-rpms.

Adding fast-datapath-for-rhel-8-x86_64-rpms to Ceph repos with a note might be a valid idea.

Personally I prefer not adding that. It is because I believe using ovs networking in
Ceph Storage nodes would introduce undesired complexity. IMHO we should have discouraged
users to use ovs in Ceph nodes.

> As I mentioned above
Sorry ignore this. I missed removing this while I updated structure of my previous comment...

No problem. You're right about fast-datapath. I think I conflated the two repos in my previous comment.

Will follow-up on this.

The issue I mentioned earlier has been fixed and now ha repo is no longer used by Ceph nodes in any case.
 https://bugzilla.redhat.com/show_bug.cgi?id=1898805

The changes are now published on the Customer Portal:
https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.1/html-single/framework_for_upgrades_13_to_16.1/index#overcloud-repositories

https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.2-beta/html-single/framework_for_upgrades_13_to_16.2/index#overcloud-repositories

Thank you

Hi Vlada,

Thank you for working on this.

The same issue exists in the installation guide, too, so I have created a separate bug for that.
If you can also fix the installation guide at this chance, that would be great.

https://bugzilla.redhat.com/show_bug.cgi?id=1989764

Hi Takashi,
Thank you for your comment.

This specific module of content "Overcloud repositories" is being reused in several guides:
* Director Installation and Usage
* Framework for Upgrades
* Release Notes

So having made this update to the module means that the change is then applied to all the docs that reuse it. There was probably a sync delay with the publishing of the updated guides, so you might have caught an older version of the Director guide and saw the repo still listed there. Apologies about that.

I can see my colleague has closed BZ#1989764 and all updated docs are now published and available on the Customer Portal.
Please let me know if you spot anything else that doesn't look right.
Many thanks!

Kind regards,
Vlada


(In reply to Takashi Kajinami from comment #13)
> Hi Vlada,
> 
> Thank you for working on this.
> 
> The same issue exists in the installation guide, too, so I have created a
> separate bug for that.
> If you can also fix the installation guide at this chance, that would be
> great.
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=1989764

Hi Vlada,

Thank you for checking it.
Oh I was not aware of that modular structure. That is very smart.

I checked the installation guide this morning (in my APAC timezone) and at that time
I saw ha repo in included in the table, but now it is completely removed.

So seems like I was looking at old content not yet synced (or maybe looking at data left in my browser cache)

No need for apologies at all. And MY apologies for making a noise here.

Thank you,
Takashi