Marcelo Vanzin created SPARK-19556:
--------------------------------------
Summary: Broadcast data is not encrypted when I/O encryption is on
Key: SPARK-19556
URL: https://issues.apache.org/jira/browse/SPARK-19556
Project: Spark
Issue Type: Bug
Components: Spark Core
Affects Versions: 2.1.0
Reporter: Marcelo Vanzin
{{TorrentBroadcast}} uses a couple of "back doors" into the block manager to
write and read data:
{code}
if (!blockManager.putBytes(pieceId, bytes, MEMORY_AND_DISK_SER,
tellMaster = true)) {
throw new SparkException(s"Failed to store $pieceId of $broadcastId in
local BlockManager")
}
{code}
{code}
bm.getLocalBytes(pieceId) match {
case Some(block) =>
blocks(pid) = block
releaseLock(pieceId)
case None =>
bm.getRemoteBytes(pieceId) match {
case Some(b) =>
if (checksumEnabled) {
val sum = calcChecksum(b.chunks(0))
if (sum != checksums(pid)) {
throw new SparkException(s"corrupt remote block $pieceId of
$broadcastId:" +
s" $sum != ${checksums(pid)}")
}
}
// We found the block from remote executors/driver's
BlockManager, so put the block
// in this executor's BlockManager.
if (!bm.putBytes(pieceId, b, StorageLevel.MEMORY_AND_DISK_SER,
tellMaster = true)) {
throw new SparkException(
s"Failed to store $pieceId of $broadcastId in local
BlockManager")
}
blocks(pid) = b
case None =>
throw new SparkException(s"Failed to get $pieceId of
$broadcastId")
}
}
{code}
The thing these block manager methods have in common is that they bypass the
encryption code; so broadcast data is stored unencrypted in the block manager,
causing unencrypted data to be written to disk if those blocks need to be
evicted from memory.
The correct fix here is actually not to change {{TorrentBroadcast}}, but to fix
the block manager so that:
- data stored in memory is not encrypted
- data written to disk is encrypted
This would simplify the code paths that use BlockManager / SerializerManager
APIs (e.g. see SPARK-19520), but requires some tricky changes inside the
BlockManager to still be able to use file channels to avoid reading whole
blocks back into memory so they can be decrypted.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
