Aut: There is insufficient memory for the Java Runtime Environment to continue

[ianmilligan1]

Running locally on an Azure 16 core, 55GB machine, Ubuntu 16; analyzing a 293GB collection that has some large WARCs (i.e. 7GB). This machine has previously happily processed ~4TB collections. There's a lot of space on the machine which is a fairly vanilla VM.

This command is used to start spark-shell:

./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --driver-memory 52G --packages "io.archivesunleashed:aut:0.12.1"

And it failed on this relatively straightforward script to count domains:

import io.archivesunleashed.spark.matchbox.{ExtractDomain, ExtractLinks, RemoveHTML, RecordLoader, WriteGEXF}
import io.archivesunleashed.spark.rdd.RecordRDD._
val r = RecordLoader.loadArchives("/data2/toronto-mayor/*.gz", sc).keepValidPages().map(r => ExtractDomain(r.getUrl)).countItems().saveAsTextFile("/data2/toronto-mayor-data/all-domains")

The error is:

OpenJDK 64-Bit Server VM warning: INFO: os::commit_memory(0x00007f62d3100000, 4703911936, 0) failed; error='Cannot allocate memory' (errno=12)
#
# There is insufficient memory for the Java Runtime Environment to continue.
# Native memory allocation (mmap) failed to map 4703911936 bytes for committing reserved memory.
# An error report file with more information is saved as:
# /home/ubuntu/aut/hs_err_pid2492.log

Error Logs

Here are the error logs: the terminal output and the hs_err_pid2492.log .

I will continue to tackle tomorrow, but any thoughts or guidance greatly appreciated.

[ruebot]

Dumping this in here for posterity.

[ianmilligan1]

fyi for me: https://www.digitalocean.com/community/tutorials/how-to-add-swap-space-on-ubuntu-16-04

[ianmilligan1]

New error when re-running with swap space (I lowered the driver memory to 45GB).

.SparkException: Job aborted due to stage failure: Task 13 in stage 0.0 failed 1 times, most recent failure: Lost task 13.0 in stage 0.0 (TID 13, localhost, executor driver): java.lang.OutOfMemoryError: Java heap space
        at java.util.Arrays.copyOf(Arrays.java:3332)
        at java.lang.StringCoding.safeTrim(StringCoding.java:89)
        at java.lang.StringCoding.access$100(StringCoding.java:50)
        at java.lang.StringCoding$StringDecoder.decode(StringCoding.java:154)
        at java.lang.StringCoding.decode(StringCoding.java:193)
        at java.lang.StringCoding.decode(StringCoding.java:254)
        at java.lang.String.<init>(String.java:546)
        at java.lang.String.<init>(String.java:566)
        at io.archivesunleashed.data.WarcRecordUtils.getWarcResponseMimeType(WarcRecordUtils.java:102)
        at io.archivesunleashed.spark.archive.io.ArchiveRecord.<init>(ArchiveRecord.scala:74)
        at io.archivesunleashed.spark.matchbox.RecordLoader$$anonfun$2.apply(RecordLoader.scala:37)
        at io.archivesunleashed.spark.matchbox.RecordLoader$$anonfun$2.apply(RecordLoader.scala:37)
        at scala.collection.Iterator$$anon$11.next(Iterator.scala:409)
        at scala.collection.Iterator$$anon$13.hasNext(Iterator.scala:462)
        at scala.collection.Iterator$$anon$13.hasNext(Iterator.scala:461)
        at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408)
        at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408)
        at org.apache.spark.util.collection.ExternalSorter.insertAll(ExternalSorter.scala:191)
        at org.apache.spark.shuffle.sort.SortShuffleWriter.write(SortShuffleWriter.scala:63)
        at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96)
        at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53)
        at org.apache.spark.scheduler.Task.run(Task.scala:99)
        at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:322)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
        at java.lang.Thread.run(Thread.java:748)

Driver stacktrace:))

[ruebot]

@ianmilligan1 what happened with the swap? Did it all get used?

[ianmilligan1]

I'm not sure, it didn't generate a nice error log this time. On the bright side, it failed sooner.

Here is the full error log dump.

[ruebot]

Can you run it again, and watch htop or something similar to see how the swap is utilized?

[ianmilligan1]

Will do!

[ianmilligan1]

OK failed after watching htop. Memory never went above the allocated amount, and swap space was essentially unused.. just suddenly and arbitrarily fails, with the same error message as above.

[ianmilligan1]

I've tried the following (and spent most of the afternoon watching logs, trying different things). In particular the following doesn't work:

• increasing swap space to 100GB; fails in same way
• using --conf spark.memory.fraction=0.4 to increase overhead room; fails in same way
• dramatically increasing partitions with -conf spark.default.parallelism=64, so does using 500; fails in same way
• trying different driver-memory configurations always seems to fail (short of getting an even-beefier machine, which I guess we could try);

We've had this issue since September 2016. https://github.com/lintool/warcbase/issues/246

[ianmilligan1]

List of failed attempts:

./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --driver-memory 45G --packages "io.archivesunleashed:aut:0.12.1"
./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --conf spark.memory.fraction=0.4 --driver-memory 45G --packages "io.archivesunleashed:aut:0.12.1"
./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --conf spark.memory.fraction=0.8 --driver-memory 45G --packages "io.archivesunleashed:aut:0.12.1"
./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --conf spark.default.parallelism=64 --driver-memory 45G --packages "io.archivesunleashed:aut:0.12.1"
./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --conf spark.default.parallelism=500 --driver-memory 45G --packages "io.archivesunleashed:aut:0.12.1"
./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --conf spark.memory.offHeap.enabled=true --conf spark.memory.offHeap.size=100G --driver-memory 45G --packages "io.archivesunleashed:aut:0.12.1"
./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --driver-memory 45G --executor-memory 10G --packages "io.archivesunleashed:aut:0.12.1"
./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --driver-memory 45G --executor-memory 45G --packages "io.archivesunleashed:aut:0.12.1"

Here's a full Spark log on the failure.

[ianmilligan1]

I got it to work!

./spark-2.1.1-bin-hadoop2.6/bin/spark-shell --master local[12] --driver-memory 45G --packages "io.archivesunleashed:aut:0.12.1"

Limiting the number of cores - default was to go to all 16 worker threads, by reducing it to 12 it worked. Going to try to run all derivatives, and if it works, will close.