Re: Bluestore onode improvements suggestion.

[Igor Fedotov <ifedotov@xxxxxxx>]

Because this ends up in a single pretty large rocksdb record which is 
expensive to modify. And the latter happens on the each write access to 
this onode.

On 5/17/2019 2:42 PM, Виталий Филиппов wrote:
> Why is it required to split blobs at all? Why not just write 4mb when 
> the client writes 4mb?
> --
> With best regards,
> Vitaliy Filippov