Ceph » mgr » Dashboard

• Backend implementation:
1. If I understand the issue correctly, we want to set the OSD flags noout, noscrub, nodeepscrub (which are already cluster-wide configurable) on per OSD level. The cluster-wide implementation is using the following command to set the flags:

   ceph osd set <flag>

   https://docs.ceph.com/docs/master/rados/operations/health-checks/#osdmap-flags
   For the configuration on per OSD level, the following command could be suitable:

   ceph osd set-group <flags> <who>

   https://docs.ceph.com/docs/master/rados/operations/health-checks/#osdmap-flags . Besides a list of flags it's possible to define a list of OSDs the flags should be applied to. The set-group command does only apply to the noout flag as it doesn't handle noscrub and nodeepscrub. I haven't seen a command to configure noscrub and nodeepscrub per OSD.
2. The OsdFlagsController:bulk_set (https://github.com/ceph/ceph/blob/master/src/pybind/mgr/dashboard/controllers/osd.py#L251) could be enhanced in order to take a list of OSDs and handle the flags accordingly.

• Frontend implementation:
  My idea would be to add another item 'Flags' (as we already have for the cluster-wide settings) to the OSD specific settings. It's possible to select one or more OSDs to configure the flags:
  
  If the user click on the 'Flags' item, a modal dialogue will open and contain three checkboxes for noout, noscrub and nodeepscrub:
  

Thanks for preparing this draft. I like your proposal! One question though: how will it be possible to see which of these flags have been enabled on a particular OSD?
Do you plan to add a new column with badges? Something else?

Lenz Grimmer wrote:

Thanks for preparing this draft. I like your proposal! One question though: how will it be possible to see which of these flags have been enabled on a particular OSD?
Do you plan to add a new column with badges? Something else?

Thank you for your comment. To be honest, I haven't thought about it yet. I would be fine with adding a column containing badges if we need the information prominently within the table view.

I'm still wondering how we want to/can "synchronize" the OSD specific with the cluster-wide flags. Let's say the 'noscrub' flag is already activated cluster-wide. Do we want to deactivate the 'noscrub' checkbox in the OSD specific settings? - Or is it possible to allow e.g. scrub for a specific OSD although 'noscrub' is set cluster-widely? I'm trying to figure out which setting overwrites another or is more important.

In relation to your question: to we then also show the cluster-wide settings as badges within the table view for every OSD? - Or do we only show the badges if the configuration is OSD specific?

There are still a lot of questions to clarify ;)

If OSD specific flags override the cluster wide flags then it should be simple to display the correct flags as badges in a column.

The only thing i'm asking myself is how to display flags if a dialog is opened after selecting multiple OSDs. In this case it is not possible to show the current values because they may differ for every OSD. Because of that there are two solutions:

• Show a dialog with unchecked flags
• Do not allow multi selection (my prefered solution)