Add PreferSameShardOnResizeWeigher.

[kpawar-sap]

This filter enable to select same host-aggregate/shard/VC for instance
resize because it could take more time to migrate the volumes over other
shards.

[joker-at-work]

Please rename to shard, as it doesn't allow to configure the aggregate and thus the more generic "PreferSameAggregate" is not telling the story correctly.

[kpawar-sap]

Please rename to shard, as it doesn't allow to configure the aggregate and thus the more generic "PreferSameAggregate" is not telling the story correctly.

Done. Tested on qa-de-1.

[fwiesel]

Otherwise, it looks good to me

[joker-at-work]

There's a test missing that actually tests the same shard part of the weigher. Please add an additional host to your tests that's in the same shard as one of the existing hosts. This host should have the same weight given by the weigher as the other host in that shard in every test.

I assume this test with then show a bug in your weigher, as the weigher returns 0.0 if the instance is not on the currently-checked host through the first if.

[joker-at-work]

Why does a new instance have _nova_check_type: ['resize'] set?

[kpawar-sap]

just to tell that even if resize is triggered with new instance it wont work since instance is new and "source_host" is missing

[joker-at-work]

This is not clear from the code, as a new instance normally wouldn't have resize set. Please add a comment that explains why, so anybody reading the code later on can understand it.

[joker-at-work]

Why do we need a set() here?
Why do we need the iterator here instead of directly using host_shard_aggr.host?
Wouldn't if instance_host in host_shard_aggr.hosts: do the same in just one line with the same readability?

[joker-at-work]

This is not clear from the code, as a new instance normally wouldn't have resize set. Please add a comment that explains why, so anybody reading the code later on can understand it.