Primary keys number metric - strange value during rebalancing

[Alexey Kuznetsov-2]

Igniters,

I see strange value of primary keys number metric.

How to reproduce:
1) Start one node with partitioned cache with 1 backup.
     Load 30M keys.
     Visor show 30M primary keys - OK so far.

2) Start second node. Rebalance started.
    And Visor shows ~15M primary keys (summed for 2 nodes) and this number
grows to 30M during rebalancing.

Looks very strange for me.

Any comments? How this metric is implemented?

See VisorCache.java line 201 about how Visor collect this metic.





--
Alexey Kuznetsov
GridGain Systems
www.gridgain.com

[yzhdanov]

I see the point. The thing is that primary keys metric makes sense only on
stable (balanced) topology.

I would suggest mentioning this in documentation and move on.

--Yakov

2015-07-03 13:32 GMT+03:00 Alexey Kuznetsov <[hidden email]>:

> Igniters,
>
> I see strange value of primary keys number metric.
>
> How to reproduce:
> 1) Start one node with partitioned cache with 1 backup.
>      Load 30M keys.
>      Visor show 30M primary keys - OK so far.
>
> 2) Start second node. Rebalance started.
>     And Visor shows ~15M primary keys (summed for 2 nodes) and this number
> grows to 30M during rebalancing.
>
> Looks very strange for me.
>
> Any comments? How this metric is implemented?
>
> See VisorCache.java line 201 about how Visor collect this metic.
>
>
>
>
>
> --
> Alexey Kuznetsov
> GridGain Systems
> www.gridgain.com
>