Cinder database for mounts to be manage with API call

Bug #1238593 reported by Dafna Ron on 2013-10-11
This bug affects 1 person
Affects Status Importance Assigned to Milestone

Bug Description

In the past few days I have been working on cinder with gluster backend.
I have encountered some major issues in how cinder works with mounts during a failure on gluster storage side.

some of the issues are nfs stale connections, cinder-volume service crashing when cannot reach a share, cinder db refresh.

speaking to eharney he said that he has been thinking that perhaps it would be a good idea to axe the shares file cinder is using and manage the mounts with API calls.

I think this would be a good idea so opening an RFE for this.

tags: added: drivers glusterfs
Avishay Traeger (avishay-il) wrote :

It seems that this may warrant a blueprint.

Ben Swartzlander (bswartz) wrote :

Since NFS drivers are not a core feature in cinder, any API to manage them would need to be an extension to cinder, and therefore not allowed to modify the DB schema. The most logical way to implement such an API would be for it to store the list of shares in a config file... which is what we already do.

My alternate suggestion is to implement config file polling in the NFS drivers, such that modifications to the shares.conf file are picked up automatically (perhaps after a short delay).

If we do that, then the only enhancement left would be to allow modification of the config file through an API rather than directly, which seems like a lot of work for relatively little gain.

Eric Harney (eharney) wrote :

I filed a blueprint yesterday to track ideas I had around this. All very theoretical still.

Eric Harney (eharney) on 2014-01-24
Changed in cinder:
importance: Undecided → Wishlist
Changed in cinder:
status: New → Won't Fix
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers