This error comes only for the Spark versions 1.4 but after the updation of Spark in Spark1.5+, you won’t get any such error with Tachyon.
Also, as your program is not clear and complete, I am guessing that in your program, if you use input.persist(OFF_HEAP), it means that the data would have two copies in Tachyon. One copy is the file itself in Tachyon, another copy in Tachyon is the OFF_HEAP store in Tachyon. Therefore, the input.persist() in your program is not recommended.
If you want to know more about Spark, then do check out this awesome video tutorial: