Comment 2 for bug 1957075

Revision history for this message
Maurice Escher (maurice-escher) wrote (last edit ):

I see there was some confusion in the meeting, I hope I can clear it up a bit:

The active replica does not have grow_shrink in this scenario, but only the passive.
And that is not, because we want that, but because that is the default working mode of snapmirror as far as I know: if you extend the size on the snapmirror source, you conveniently don't have to manually extend the size on the target, too. I'm not sure if this manual size change would even be possible, maybe with xdp, but most likely not with dp snapmirrors.

So the active replica will have the correct (manila = source of truth) size. If it is pretty much full the passive replica may already grow bigger depending on the autosize thresholds.

But now comes the replica promote: the old active replica is now passive and keeps its autosize none setting, and the old passive is now active and keeps its autosize grow_shrink setting.

Now the active one can grow out of control (bad because it is bigger than what is known to manila, bigger than what is billed for), and the new snapmirror target can not follow to grow automatically (bad as well, because sync will not work)