[SPARK-18274] Memory leak in PySpark StringIndexer

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.5.2, 1.6.3, 2.0.1, 2.0.2, 2.1.0
• Fix Version/s: 2.0.3, 2.1.0, 2.2.0
• Component/s: ML, PySpark
• Target Version/s: 2.0.3, 2.1.0
• Assignee: Sandeep Singh
• Reporter: Jonas Amrich
• Shepherd: Joseph K. Bradley
• Created: 04/Nov/16 15:31
• Updated: 08/Dec/16 07:29
• Resolved: 01/Dec/16 21:23

Description

StringIndexerModel won't get collected by GC in Java even when deleted in Python. It can be reproduced by this code, which fails after couple of iterations (around 7 if you set driver memory to 600MB):

import random, string
from pyspark.ml.feature import StringIndexer

l = [(''.join(random.choice(string.ascii_uppercase) for _ in range(10)), ) for _ in range(int(7e5))]  # 700000 random strings of 10 characters
df = spark.createDataFrame(l, ['string'])

for i in range(50):
    indexer = StringIndexer(inputCol='string', outputCol='index')
    indexer.fit(df)

Explicit call to Python GC fixes the issue - following code runs fine:

for i in range(50):
    indexer = StringIndexer(inputCol='string', outputCol='index')
    indexer.fit(df)
    gc.collect()

The issue is similar to SPARK-6194 and can be probably fixed by calling jvm detach in model's destructor. This is implemented in pyspark.mlib.common.JavaModelWrapper but missing in pyspark.ml.wrapper.JavaWrapper. Other models in ml package may also be affected by this memory leak.

Issue Links

• relates to: SPARK-18630 PySpark ML memory leak
• links to: [Github] Pull Request #15843 (techaddict)