Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Halt Spark executor when encountering unrecoverable CUDA errors #5350

Merged
merged 11 commits into from
Jun 13, 2022
Merged

Halt Spark executor when encountering unrecoverable CUDA errors #5350

Changes from 10 commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
6f28f52
refine the handling of task failure reason
sperlingxx Apr 28, 2022
30b042d
fix style
sperlingxx Apr 28, 2022
6afd0cc
update
sperlingxx Apr 29, 2022
7adea66
Merge remote-tracking branch 'origin/branch-22.06' into recognize_cud…
sperlingxx May 11, 2022
3ae85a9
fix
sperlingxx May 12, 2022
4492061
fix
sperlingxx May 12, 2022
e083451
Merge remote-tracking branch 'origin/branch-22.08' into recognize_cud…
sperlingxx Jun 6, 2022
18f0dac
update
sperlingxx Jun 6, 2022
36a73f7
Update sql-plugin/src/main/scala/com/nvidia/spark/rapids/Plugin.scala
sperlingxx Jun 8, 2022
7a9c6b6
update
sperlingxx Jun 8, 2022
464db65
Update sql-plugin/src/main/scala/com/nvidia/spark/rapids/Plugin.scala
sperlingxx Jun 8, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 14 additions & 11 deletions sql-plugin/src/main/scala/com/nvidia/spark/rapids/Plugin.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ import scala.collection.mutable.{Map => MutableMap}
import scala.util.Try
import scala.util.matching.Regex

import ai.rapids.cudf.{CudaException, CudaFatalException, CudfException}
import com.nvidia.spark.rapids.python.PythonWorkerSemaphore

import org.apache.spark.{ExceptionFailure, SparkConf, SparkContext, TaskFailedReason}
Expand Down Expand Up @@ -288,19 +289,21 @@ class RapidsExecutorPlugin extends ExecutorPlugin with Logging {
override def onTaskFailed(failureReason: TaskFailedReason): Unit = {
failureReason match {
case ef: ExceptionFailure =>
val unrecoverableErrors = Seq("cudaErrorIllegalAddress", "cudaErrorLaunchTimeout",
"cudaErrorHardwareStackError", "cudaErrorIllegalInstruction",
"cudaErrorMisalignedAddress", "cudaErrorInvalidAddressSpace", "cudaErrorInvalidPc",
"cudaErrorLaunchFailure", "cudaErrorExternalDevice", "cudaErrorUnknown",
"cudaErrorECCUncorrectable")
if (unrecoverableErrors.exists(ef.description.contains(_)) ||
unrecoverableErrors.exists(ef.toErrorString.contains(_))) {
logError("Stopping the Executor based on exception being a fatal CUDA error: " +
s"${ef.toErrorString}")
System.exit(20)
ef.exception match {
case _: CudaFatalException =>
logError("Stopping the Executor based on exception being a fatal CUDA error: " +
s"${ef.toErrorString}")
System.exit(20)
case _: CudaException =>
logDebug(s"Executor onTaskFailed because of a non-fatal CUDA error: " +
s"${ef.toErrorString}")
case _: CudfException =>
logDebug(s"Executor onTaskFailed because of a CUDF error: ${ef.toErrorString}")
sperlingxx marked this conversation as resolved.
Show resolved Hide resolved
case _ =>
logDebug(s"Executor onTaskFailed: ${ef.toErrorString}")
}
case other =>
logDebug(s"Executor onTaskFailed not a CUDA fatal error: ${other.toString}")
logDebug(s"Executor onTaskFailed: ${other.toString}")
}
}
}
Expand Down