[heketi-devel] Question regarding heketi topology and clusterid/nodeid

[Raghavendra Talur]

On Fri, Apr 12, 2019 at 9:52 AM Grumboeck Johannes (POI - AT/Salzburg)
<johannes.grumboeck at porscheinformatik.at> wrote:
>
> Hi,
>
>
>
> Thanks for your answer. I’ve got one more question.
>
>
>
> > > Are clusterid and nodeid something heketi uses only internally and are not
> > > directly connected to something within glusterfs?
> > Yes, in glusterd there is no concept of clusterid. It is an id that heketi creates to refer to a particular cluster.
> > Nodeid: glusterd generates a node id for all peers in the cluster but the nodeid in heketi topology info is "local" only to heketi as it has to create the nodeid before the peer probe command.
>
> Does this mean that when I could change clusterid and nodeids in a heketi-topology file,
>
> reload it into heketi db and, as long as all references are fine within the topology file, all volumes and bricks should be safe?

Yes replacing the ids is fine.
>
>
>
> I’m asking because I’m thinking about writing a script to generate a heketi-topology out of a gluster with all volumes/bricks in a safe state
>
> to recreate a corrupted heketi.db file.

A word of warning, it is not trivial but it is doable. I suggest that
you build a heketi db representation in json like the
"heketi-cli db export" gives instead of building the topology file
representation.

If you have the json then creating a db out of it is easy using
"heketi-cli db import" command. One added work would be to generate
UUID for all the resources.

Let me know what you think. We had plans to doing a similar feature
but never prioritized.

Thanks,
Talur

>
>
>
> Greetings,
>
> Johannes
>
>
>
> Internal