[
https://issues.apache.org/jira/browse/SPARK-26549?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yuanjian Li updated SPARK-26549:
--------------------------------
Description:
During [the follow-up
work|https://github.com/apache/spark/pull/23435#issuecomment-451079886] for
PySpark worker reuse scenario, we found that the worker reuse takes no effect
for `sc.parallelize(xrange(...))`.
It happened because the specialize rdd.parallelize for xrange generated data by
xrange, which don't need the passed-in iterator. But this will break the end of
stream checking in python worker and finally cause worker reuse take no effect.
Relative code block list below:
Current specialize logic of xrange don't need the passed-in iterator,
context.py:
{code:java}
if isinstance(c, xrange):
...
def f(split, iterator):
return xrange(getStart(split), getStart(split + 1), step)
...
return self.parallelize([], numSlices).mapPartitionsWithIndex(f)
{code}
We got an unexpected value -1 which refers to END_OF_DATA_SECTION while check
end of stream. See the code in worker.py:
{code:java}
# check end of stream
if read_int(infile) == SpecialLengths.END_OF_STREAM:
write_int(SpecialLengths.END_OF_STREAM, outfile)
else:
# write a different value to tell JVM to not reuse this worker
write_int(SpecialLengths.END_OF_DATA_SECTION, outfile)
sys.exit(-1)
{code}
The code works well for parallelize(range) because the END_OF_DATA_SECTION has
been handled during load iterator from the socket stream, see the code in
FramedSerializer:
{code:java}
def load_stream(self, stream):
while True:
try:
yield self._read_with_length(stream)
except EOFError:
return
...
def _read_with_length(self, stream):
length = read_int(stream)
if length == SpecialLengths.END_OF_DATA_SECTION:
raise EOFError #END_OF_DATA_SECTION raised EOF here and catched in
load_stream
elif length == SpecialLengths.NULL:
return None
obj = stream.read(length)
if len(obj) < length:
raise EOFError
return self.loads(obj)
{code}
was:
During [the follow-up
work|https://github.com/apache/spark/pull/23435#issuecomment-451079886] for
PySpark worker reuse scenario, we found that the worker reuse takes no effect
for `sc.parallelize(xrange(...))`.
It happened because the specialize logic generated data by xrange, which don't
need the passed-in iterator. But this will break the end of stream checking in
python worker and finally cause worker reuse take no effect.
Relative code block list below:
Current specialize logic of xrange don't need the passed-in iterator,
context.py:
{code}
if isinstance(c, xrange):
...
def f(split, iterator):
return xrange(getStart(split), getStart(split + 1), step)
...
return self.parallelize([], numSlices).mapPartitionsWithIndex(f)
{code}
We got an unexpected value -1 which refers to END_OF_DATA_SECTION while check
end of stream. See the code in worker.py:
{code}
# check end of stream
if read_int(infile) == SpecialLengths.END_OF_STREAM:
write_int(SpecialLengths.END_OF_STREAM, outfile)
else:
# write a different value to tell JVM to not reuse this worker
write_int(SpecialLengths.END_OF_DATA_SECTION, outfile)
sys.exit(-1)
{code}
The code works well for parallelize(range) because the END_OF_DATA_SECTION has
been handled during load iterator from the socket stream, see the code in
FramedSerializer:
{code}
def load_stream(self, stream):
while True:
try:
yield self._read_with_length(stream)
except EOFError:
return
...
def _read_with_length(self, stream):
length = read_int(stream)
if length == SpecialLengths.END_OF_DATA_SECTION:
raise EOFError #END_OF_DATA_SECTION raised EOF here and catched in
load_stream
elif length == SpecialLengths.NULL:
return None
obj = stream.read(length)
if len(obj) < length:
raise EOFError
return self.loads(obj)
{code}
> PySpark worker reuse take no effect for parallelize xrange
> ----------------------------------------------------------
>
> Key: SPARK-26549
> URL: https://issues.apache.org/jira/browse/SPARK-26549
> Project: Spark
> Issue Type: Bug
> Components: PySpark
> Affects Versions: 3.0.0
> Reporter: Yuanjian Li
> Priority: Major
>
> During [the follow-up
> work|https://github.com/apache/spark/pull/23435#issuecomment-451079886] for
> PySpark worker reuse scenario, we found that the worker reuse takes no effect
> for `sc.parallelize(xrange(...))`.
> It happened because the specialize rdd.parallelize for xrange generated data
> by xrange, which don't need the passed-in iterator. But this will break the
> end of stream checking in python worker and finally cause worker reuse take
> no effect.
> Relative code block list below:
> Current specialize logic of xrange don't need the passed-in iterator,
> context.py:
> {code:java}
> if isinstance(c, xrange):
> ...
> def f(split, iterator):
> return xrange(getStart(split), getStart(split + 1), step)
> ...
> return self.parallelize([], numSlices).mapPartitionsWithIndex(f)
> {code}
> We got an unexpected value -1 which refers to END_OF_DATA_SECTION while check
> end of stream. See the code in worker.py:
> {code:java}
> # check end of stream
> if read_int(infile) == SpecialLengths.END_OF_STREAM:
> write_int(SpecialLengths.END_OF_STREAM, outfile)
> else:
> # write a different value to tell JVM to not reuse this worker
> write_int(SpecialLengths.END_OF_DATA_SECTION, outfile)
> sys.exit(-1)
> {code}
> The code works well for parallelize(range) because the END_OF_DATA_SECTION
> has been handled during load iterator from the socket stream, see the code in
> FramedSerializer:
> {code:java}
> def load_stream(self, stream):
> while True:
> try:
> yield self._read_with_length(stream)
> except EOFError:
> return
> ...
> def _read_with_length(self, stream):
> length = read_int(stream)
> if length == SpecialLengths.END_OF_DATA_SECTION:
> raise EOFError #END_OF_DATA_SECTION raised EOF here and catched in
> load_stream
> elif length == SpecialLengths.NULL:
> return None
> obj = stream.read(length)
> if len(obj) < length:
> raise EOFError
> return self.loads(obj)
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
