online schema changes with deletion.
architecture as detailed explained here: https://kb.wixpress.com/x/8pocBQ

Where clause condition will always receive the default 1=1.
in case we would like to delete table data we will use the where clause as in a "select" meaning we will put the data we want to save in the where clause.
to include values on the where clause - use the condition as it is. for exclude values please put "not" before.

Please note : 
Where clause condition will be enforced only from the source table and not from the binlog. meaning if I put a where clause of id < 100, and id # 2 was updated \ inserted, it will be in the new table because this DML will be taken from the binlog. this is a feature since this is newly update\insert otherwise it should be inserted into the binlogs as a deleted row.
If you are using one filter operation with in  , please don't put more than 5 values. otherwise please use the sub-select option.
For hash \ string it is better to use the sub-select option, avoiding string manipulation issues in the middle (Jenkins, ansible, Casper). 
Where clause can not be edited in the middle of ghost run.