Capacity data On ES

I have 3 node in my cluster. I using ES version 2.1.1
I have 14 Indices with 673,641,059 document. with 1 Share và 1 Rep
I check on Marvel, Data is 357GB.
But i check on disk, Data used is 532GB???
Why ?. Idon't understand, please explain for me. Thanks you!

ES-01

/dev/sda1             485M   33M  427M   8% /boot
/dev/sda3              57G   26G   29G  48% /data/data1
/dev/sdb1             197G  107G   81G  57% /data/data2

ES-02

tmpfs                 7.8G   88K  7.8G   1% /dev/shm
/dev/sda1             485M   33M  427M   8% /boot
/dev/sda3             163G  135G   20G  88% /data/data1

ES-03

/dev/sda1             485M   33M  427M   8% /boot
/dev/sda3              50G   46G  796M  99% /data/data1
/dev/sdb1             266G  224G   30G  89% /data/data2

Maybe deleted documents?
It could also be things like doc values.

Yes, before, i delete all data and create new indices. Is that make problem?