This is more general than simply passing a string name and leaves more room for performance optimizations.
Note that this is technically an API breaking change in the following two ways:
1. The shuffle serializer specification in ShuffleDependency now require an object instead of a String (of the class name), but I suspect nobody else in this world has used this API other than me in GraphX and Shark.
2. Serializer's in Spark from now on are required to be serializable.
Author: Reynold Xin <rxin@apache.org>
Closes#149 from rxin/serializer and squashes the following commits:
5acaccd [Reynold Xin] Properly call serializer's constructors.
2a8d75a [Reynold Xin] Added more documentation for the serializer option in ShuffleDependency.
7420185 [Reynold Xin] Allow user to pass Serializer object instead of class name for shuffle.
Author: liguoqiang <liguoqiang@rd.tuan800.com>
Closes#44 from witgo/SPARK-1149 and squashes the following commits:
3dcdcaf [liguoqiang] Merge branch 'master' into SPARK-1149
8425395 [liguoqiang] Merge remote-tracking branch 'upstream/master' into SPARK-1149
3dad595 [liguoqiang] review comment
e3e56aa [liguoqiang] Merge branch 'master' into SPARK-1149
b0d5c07 [liguoqiang] review comment
d0a6005 [liguoqiang] review comment
3395ee7 [liguoqiang] Merge remote-tracking branch 'upstream/master' into SPARK-1149
ac006a3 [liguoqiang] code Formatting
3feb3a8 [liguoqiang] Merge branch 'master' into SPARK-1149
adc443e [liguoqiang] partitions check bugfix
928e1e3 [liguoqiang] Added a unit test for PairRDDFunctions.lookup with bad partitioner
db6ecc5 [liguoqiang] Merge branch 'master' into SPARK-1149
1e3331e [liguoqiang] Merge branch 'master' into SPARK-1149
3348619 [liguoqiang] Optimize performance for partitions check
61e5a87 [liguoqiang] Merge branch 'master' into SPARK-1149
e68210a [liguoqiang] add partition index check to submitJob
3a65903 [liguoqiang] make the code more readable
6bb725e [liguoqiang] fix #SPARK-1149 Bad partitioners can cause Spark to hang
SPARK-1194: https://spark-project.atlassian.net/browse/SPARK-1194
In the current implementation, when selecting candidate blocks to be swapped out, once we find a block from the same RDD that the block to be stored belongs to, cache eviction fails and aborts.
In this PR, we keep selecting blocks *not* from the RDD that the block to be stored belongs to until either enough free space can be ensured (cache eviction succeeds) or all such blocks are checked (cache eviction fails).
Author: Cheng Lian <lian.cs.zju@gmail.com>
Closes#96 from liancheng/fix-spark-1194 and squashes the following commits:
2524ab9 [Cheng Lian] Added regression test case for SPARK-1194
6e40c22 [Cheng Lian] Remove redundant comments
40cdcb2 [Cheng Lian] Bug fix, and addressed PR comments from @mridulm
62c92ac [Cheng Lian] Fixed SPARK-1194 https://spark-project.atlassian.net/browse/SPARK-1194
Author: Prashant Sharma <prashant.s@imaginea.com>
Author: Prashant Sharma <scrapcodes@gmail.com>
Closes#80 from ScrapCodes/SPARK-1165/RDD.intersection and squashes the following commits:
9b015e9 [Prashant Sharma] Added a note, shuffle is required for intersection.
1fea813 [Prashant Sharma] correct the lines wrapping
d0c71f3 [Prashant Sharma] SPARK-1165 RDD.intersection in java
d6effee [Prashant Sharma] SPARK-1165 Implemented RDD.intersection in python.
Hadoop uses the config mapreduce.map.input.file to indicate the input filename to the map when the input split is of type FileSplit. Some of the hadoop input and output formats set or use this config. This config can also be used by user code.
PipedRDD runs an external process and the configs aren't available to that process. Hadoop Streaming does something very similar and the way they make configs available is exporting them into the environment replacing '.' with '_'. Spark should also export this variable when launching the pipe command so the user code has access to that config.
Note that the config mapreduce.map.input.file is the new one, the old one which is deprecated but not yet removed is map.input.file. So we should handle both.
Perhaps it would be better to abstract this out somehow so it goes into the HadoopParition code?
Author: Thomas Graves <tgraves@apache.org>
Closes#94 from tgravescs/map_input_file and squashes the following commits:
cc97a6a [Thomas Graves] Update test to check for existence of command, add a getPipeEnvVars function to HadoopRDD
e3401dc [Thomas Graves] Merge remote-tracking branch 'upstream/master' into map_input_file
2ba805e [Thomas Graves] set map_input_file environment variable in PipedRDD
This patch changes "yarn-standalone" to "yarn-cluster" (but still supports the former). It also cleans up the Running on YARN docs and adds a section on how to view logs.
Author: Sandy Ryza <sandy@cloudera.com>
Closes#95 from sryza/sandy-spark-1197 and squashes the following commits:
563ef3a [Sandy Ryza] Review feedback
6ad06d4 [Sandy Ryza] Change yarn-standalone to yarn-cluster and fix up running on YARN docs
resubmit pull request. was https://github.com/apache/incubator-spark/pull/332.
Author: Thomas Graves <tgraves@apache.org>
Closes#33 from tgravescs/security-branch-0.9-with-client-rebase and squashes the following commits:
dfe3918 [Thomas Graves] Fix merge conflict since startUserClass now using runAsUser
05eebed [Thomas Graves] Fix dependency lost in upmerge
d1040ec [Thomas Graves] Fix up various imports
05ff5e0 [Thomas Graves] Fix up imports after upmerging to master
ac046b3 [Thomas Graves] Merge remote-tracking branch 'upstream/master' into security-branch-0.9-with-client-rebase
13733e1 [Thomas Graves] Pass securityManager and SparkConf around where we can. Switch to use sparkConf for reading config whereever possible. Added ConnectionManagerSuite unit tests.
4a57acc [Thomas Graves] Change UI createHandler routines to createServlet since they now return servlets
2f77147 [Thomas Graves] Rework from comments
50dd9f2 [Thomas Graves] fix header in SecurityManager
ecbfb65 [Thomas Graves] Fix spacing and formatting
b514bec [Thomas Graves] Fix reference to config
ed3d1c1 [Thomas Graves] Add security.md
6f7ddf3 [Thomas Graves] Convert SaslClient and SaslServer to scala, change spark.authenticate.ui to spark.ui.acls.enable, and fix up various other things from review comments
2d9e23e [Thomas Graves] Merge remote-tracking branch 'upstream/master' into security-branch-0.9-with-client-rebase_rework
5721c5a [Thomas Graves] update AkkaUtilsSuite test for the actorSelection changes, fix typos based on comments, and remove extra lines I missed in rebase from AkkaUtils
f351763 [Thomas Graves] Add Security to Spark - Akka, Http, ConnectionManager, UI to use servlets
This is a port of a pull request original targeted at incubator-spark: https://github.com/apache/incubator-spark/pull/180
Essentially if a user returns a generative iterator (from a flatMap operation), when trying to persist the data, Spark would first unroll the iterator into an ArrayBuffer, and then try to figure out if it could store the data. In cases where the user provided an iterator that generated more data then available memory, this would case a crash. With this patch, if the user requests a persist with a 'StorageLevel.DISK_ONLY', the iterator will be unrolled as it is inputed into the serializer.
To do this, two changes where made:
1) The type of the 'values' argument in the putValues method of the BlockStore interface was changed from ArrayBuffer to Iterator (and all code interfacing with this method was modified to connect correctly.
2) The JavaSerializer now calls the ObjectOutputStream 'reset' method every 1000 objects. This was done because the ObjectOutputStream caches objects (thus preventing them from being GC'd) to write more compact serialization. If reset is never called, eventually the memory fills up, if it is called too often then the serialization streams become much larger because of redundant class descriptions.
Author: Kyle Ellrott <kellrott@gmail.com>
Closes#50 from kellrott/iterator-to-disk and squashes the following commits:
9ef7cb8 [Kyle Ellrott] Fixing formatting issues.
60e0c57 [Kyle Ellrott] Fixing issues (formatting, variable names, etc.) from review comments
8aa31cd [Kyle Ellrott] Merge ../incubator-spark into iterator-to-disk
33ac390 [Kyle Ellrott] Merge branch 'iterator-to-disk' of github.com:kellrott/incubator-spark into iterator-to-disk
2f684ea [Kyle Ellrott] Refactoring the BlockManager to replace the Either[Either[A,B]] usage. Now using trait 'Values'. Also modified BlockStore.putBytes call to return PutResult, so that it behaves like putValues.
f70d069 [Kyle Ellrott] Adding docs for spark.serializer.objectStreamReset configuration
7ccc74b [Kyle Ellrott] Moving the 'LargeIteratorSuite' to simply test persistance of iterators. It doesn't try to invoke a OOM error any more
16a4cea [Kyle Ellrott] Streamlined the LargeIteratorSuite unit test. It should now run in ~25 seconds. Confirmed that it still crashes an unpatched copy of Spark.
c2fb430 [Kyle Ellrott] Removing more un-needed array-buffer to iterator conversions
627a8b7 [Kyle Ellrott] Wrapping a few long lines
0f28ec7 [Kyle Ellrott] Adding second putValues to BlockStore interface that accepts an ArrayBuffer (rather then an Iterator). This will allow BlockStores to have slightly different behaviors dependent on whether they get an Iterator or ArrayBuffer. In the case of the MemoryStore, it needs to duplicate and cache an Iterator into an ArrayBuffer, but if handed a ArrayBuffer, it can skip the duplication.
656c33e [Kyle Ellrott] Fixing the JavaSerializer to read from the SparkConf rather then the System property.
8644ee8 [Kyle Ellrott] Merge branch 'master' into iterator-to-disk
00c98e0 [Kyle Ellrott] Making the Java ObjectStreamSerializer reset rate configurable by the system variable 'spark.serializer.objectStreamReset', default is not 10000.
40fe1d7 [Kyle Ellrott] Removing rouge space
31fe08e [Kyle Ellrott] Removing un-needed semi-colons
9df0276 [Kyle Ellrott] Added check to make sure that streamed-to-dist RDD actually returns good data in the LargeIteratorSuite
a6424ba [Kyle Ellrott] Wrapping long line
2eeda75 [Kyle Ellrott] Fixing dumb mistake ("||" instead of "&&")
0e6f808 [Kyle Ellrott] Deleting temp output directory when done
95c7f67 [Kyle Ellrott] Simplifying StorageLevel checks
56f71cd [Kyle Ellrott] Merge branch 'master' into iterator-to-disk
44ec35a [Kyle Ellrott] Adding some comments.
5eb2b7e [Kyle Ellrott] Changing the JavaSerializer reset to occur every 1000 objects.
f403826 [Kyle Ellrott] Merge branch 'master' into iterator-to-disk
81d670c [Kyle Ellrott] Adding unit test for straight to disk iterator methods.
d32992f [Kyle Ellrott] Merge remote-tracking branch 'origin/master' into iterator-to-disk
cac1fad [Kyle Ellrott] Fixing MemoryStore, so that it converts incoming iterators to ArrayBuffer objects. This was previously done higher up the stack.
efe1102 [Kyle Ellrott] Changing CacheManager and BlockManager to pass iterators directly to the serializer when a 'DISK_ONLY' persist is called. This is in response to SPARK-942.
Author: Prashant Sharma <prashant.s@imaginea.com>
Author: Patrick Wendell <pwendell@gmail.com>
Closes#17 from ScrapCodes/java8-lambdas and squashes the following commits:
95850e6 [Patrick Wendell] Some doc improvements and build changes to the Java 8 patch.
85a954e [Prashant Sharma] Nit. import orderings.
673f7ac [Prashant Sharma] Added support for -java-home as well
80a13e8 [Prashant Sharma] Used fake class tag syntax
26eb3f6 [Prashant Sharma] Patrick's comments on PR.
35d8d79 [Prashant Sharma] Specified java 8 building in the docs
31d4cd6 [Prashant Sharma] Maven build to support -Pjava8-tests flag.
4ab87d3 [Prashant Sharma] Review feedback on the pr
c33dc2c [Prashant Sharma] SPARK-964, Java 8 API Support.
Lookup didn't have a unit test. Added two tests, one for with a partitioner, and one for without.
Author: Bryn Keller <bryn.keller@intel.com>
Closes#36 from xoltar/lookup and squashes the following commits:
3bc0d44 [Bryn Keller] Added a unit test for PairRDDFunctions.lookup
This metric is confusing: it adds up all of the time to fetch
shuffle inputs, but fetches often happen in parallel, so
remoteFetchTime can be much longer than the task execution time.
@squito it looks like you added this metric -- do you have a use case for it?
cc @shivaram -- I know you've looked at the shuffle performance a lot so chime in here if this metric has turned out to be useful for you!
Author: Kay Ousterhout <kayousterhout@gmail.com>
Closes#62 from kayousterhout/remove_fetch_variable and squashes the following commits:
43341eb [Kay Ousterhout] Remote the remoteFetchTime metric.
Thanks for Diana Carroll to report this issue (https://spark-project.atlassian.net/browse/SPARK-1100)
the current saveAsTextFile/SequenceFile will overwrite the output directory silently if the directory already exists, this behaviour is not desirable because
overwriting the data silently is not user-friendly
if the partition number of two writing operation changed, then the output directory will contain the results generated by two runnings
My fix includes:
add some new APIs with a flag for users to define whether he/she wants to overwrite the directory:
if the flag is set to true, then the output directory is deleted first and then written into the new data to prevent the output directory contains results from multiple rounds of running;
if the flag is set to false, Spark will throw an exception if the output directory already exists
changed JavaAPI part
default behaviour is overwriting
Two questions
should we deprecate the old APIs without such a flag?
I noticed that Spark Streaming also called these APIs, I thought we don't need to change the related part in streaming? @tdas
Author: CodingCat <zhunansjtu@gmail.com>
Closes#11 from CodingCat/SPARK-1100 and squashes the following commits:
6a4e3a3 [CodingCat] code clean
ef2d43f [CodingCat] add new test cases and code clean
ac63136 [CodingCat] checkOutputSpecs not applicable to FSOutputFormat
ec490e8 [CodingCat] prevent Spark from overwriting directory silently and leaving dirty directory
This commit randomizes the order of resource offers to avoid scheduling
all tasks on the same small set of machines.
This is a much simpler solution to SPARK-979 than #7.
Author: Kay Ousterhout <kayousterhout@gmail.com>
Closes#27 from kayousterhout/randomize and squashes the following commits:
435d817 [Kay Ousterhout] [SPARK-979] Randomize order of offers.
(Ported from https://github.com/apache/incubator-spark/pull/637 )
Author: Sean Owen <sowen@cloudera.com>
Closes#31 from srowen/SPARK-1084.1 and squashes the following commits:
6c4a32c [Sean Owen] Suppress warnings about legitimate unchecked array creations, or change code to avoid it
f35b833 [Sean Owen] Fix two misc javadoc problems
254e8ef [Sean Owen] Fix one new style error introduced in scaladoc warning commit
5b2fce2 [Sean Owen] Fix scaladoc invocation warning, and enable javac warnings properly, with plugin config updates
007762b [Sean Owen] Remove dead scaladoc links
b8ff8cb [Sean Owen] Replace deprecated Ant <tasks> with <target>
If the seed is zero, XORShift generates all zeros, which would create unexpected result.
JIRA: https://spark-project.atlassian.net/browse/SPARK-1129
Author: Xiangrui Meng <meng@databricks.com>
Closes#645 from mengxr/xor and squashes the following commits:
1b086ab [Xiangrui Meng] use MurmurHash3 to set seed in XORShiftRandom
45c6f16 [Xiangrui Meng] minor style change
51f4050 [Xiangrui Meng] use a predefined seed when seed is zero in XORShiftRandom
ClusterScheduler was renamed to TaskSchedulerImpl; this commit
updates comments and tests accordingly.
Author: Kay Ousterhout <kayousterhout@gmail.com>
Closes#9 from kayousterhout/cluster_scheduler_death and squashes the following commits:
d6fd119 [Kay Ousterhout] Remove references to ClusterScheduler.
The aim of the Json4s project is to provide a common API for
Scala JSON libraries. It is Apache-licensed, easier for
downstream distributions to package, and mostly API-compatible
with lift-json. Furthermore, the Jackson-backed implementation
parses faster than lift-json on all but the smallest inputs.
Author: William Benton <willb@redhat.com>
Closes#582 from willb/json4s and squashes the following commits:
7ca62c4 [William Benton] Replace lift-json with json4s-jackson.
[SPARK-1108] This allows us to use, e.g. HBase's TableOutputFormat with PairRDDFunctions.saveAsNewAPIHadoopFile, which otherwise would throw NullPointerException because the output table name hasn't been configured.
Note this bug also affects branch-0.9
Author: Bryn Keller <bryn.keller@intel.com>
Closes#638 from xoltar/SPARK-1108 and squashes the following commits:
7e94e7d [Bryn Keller] Import, comment, and format cleanup per code review
7cbcaa1 [Bryn Keller] For outputformats that are Configurable, call setConf before sending data to them. This allows us to use, e.g. HBase TableOutputFormat, which otherwise would throw NullPointerException because the output table name hasn't been configured
In the previous code, if you had a failing map stage and then tried to
run reduce stages on it repeatedly, the first reduce stage would fail
correctly, but the later ones would mistakenly believe that all map
outputs are available and start failing infinitely with fetch failures
from "null".
These classes can't be migrated:
StorageLevels: impossible to create static fields in Scala
JavaSparkContextVarargsWorkaround: incompatible varargs
JavaAPISuite: should test Java APIs in pure Java (for sanity)
Author: Punya Biswal <pbiswal@palantir.com>
Closes#605 from punya/move-java-sources and squashes the following commits:
25b00b2 [Punya Biswal] Remove redundant type param; reformat
853da46 [Punya Biswal] Use factory method rather than constructor
e5d53d9 [Punya Biswal] Migrate Java code to Scala or move it to src/main/java
The original poster of this bug is @guojc, who opened a PR that preceded this one at https://github.com/apache/incubator-spark/pull/612.
ExternalAppendOnlyMap uses key hash code to order the buffer streams from which spilled files are read back into memory. When a buffer stream is empty, the default hash code for that stream is equal to Int.MaxValue. This is, however, a perfectly legitimate candidate for a key hash code. When reading from a spilled map containing such a key, a hash collision may occur, in which case we attempt to read from an empty stream and throw NoSuchElementException.
The fix is to maintain the invariant that empty buffer streams are never added back to the merge queue to be considered. This guarantees that we never read from an empty buffer stream, ever again.
This PR also includes two new tests for hash collisions.
Author: Andrew Or <andrewor14@gmail.com>
Closes#624 from andrewor14/spilling-bug and squashes the following commits:
9e7263d [Andrew Or] Slightly optimize next()
2037ae2 [Andrew Or] Move a few comments around...
cf95942 [Andrew Or] Remove default value of Int.MaxValue for minKeyHash
c11f03b [Andrew Or] Fix Int.MaxValue hash collision bug in ExternalAppendOnlyMap
21c1a39 [Andrew Or] Add hash collision tests to ExternalAppendOnlyMapSuite
Fixes an error where HDFS URL's cause an exception. Should be merged into master and 0.9.
Author: Patrick Wendell <pwendell@gmail.com>
Closes#625 from pwendell/url-validation and squashes the following commits:
d14bfe3 [Patrick Wendell] SPARK-1111: URL Validation Throws Error for HDFS URL's
Optimized imports and arranged according to scala style guide @
https://cwiki.apache.org/confluence/display/SPARK/Spark+Code+Style+Guide#SparkCodeStyleGuide-Imports
Author: NirmalReddy <nirmal.reddy@imaginea.com>
Author: NirmalReddy <nirmal_reddy2000@yahoo.com>
Closes#613 from NirmalReddy/opt-imports and squashes the following commits:
578b4f5 [NirmalReddy] imported java.lang.Double as JDouble
a2cbcc5 [NirmalReddy] addressed the comments
776d664 [NirmalReddy] Optimized imports in core
SPARK-1076: zipWithIndex and zipWithUniqueId to RDD
Assign ranks to an ordered or unordered data set is a common operation. This could be done by first counting records in each partition and then assign ranks in parallel.
The purpose of assigning ranks to an unordered set is usually to get a unique id for each item, e.g., to map feature names to feature indices. In such cases, the assignment could be done without counting records, saving one spark job.
https://spark-project.atlassian.net/browse/SPARK-1076
== update ==
Because assigning ranks is very similar to Scala's zipWithIndex, I changed the method name to zipWithIndex and put the index in the value field.
Author: Xiangrui Meng <meng@databricks.com>
Closes#578 and squashes the following commits:
52a05e1 [Xiangrui Meng] changed assignRanks to zipWithIndex changed assignUniqueIds to zipWithUniqueId minor updates
756881c [Xiangrui Meng] simplified RankedRDD by implementing assignUniqueIds separately moved couting iterator size to Utils do not count items in the last partition and skip counting if there is only one partition
630868c [Xiangrui Meng] newline
21b434b [Xiangrui Meng] add assignRanks and assignUniqueIds to RDD
SPARK-1072 Use binary search when needed in RangePartioner
Author: Holden Karau <holden@pigscanfly.ca>
Closes#571 and squashes the following commits:
f31a2e1 [Holden Karau] Swith to using CollectionsUtils in Partitioner
4c7a0c3 [Holden Karau] Add CollectionsUtil as suggested by aarondav
7099962 [Holden Karau] Add the binary search to only init once
1bef01d [Holden Karau] CR feedback
a21e097 [Holden Karau] Use binary search if we have more than 1000 elements inside of RangePartitioner
[SPARK-1038] Add more fields in JsonProtocol and add tests that verify the JSON itself
This is a PR for SPARK-1038. Two major changes:
1 add some fields to JsonProtocol which is new and important to standalone-related data structures
2 Use Diff in liftweb.json to verity the stringified Json output for detecting someone mod type T to Option[T]
Author: qqsun8819 <jin.oyj@alibaba-inc.com>
Closes#551 and squashes the following commits:
fdf0b4e [qqsun8819] [SPARK-1038] 1. Change code style for more readable according to rxin review 2. change submitdate hard-coded string to a date object toString for more complexiblity
095a26f [qqsun8819] [SPARK-1038] mod according to review of pwendel, use hard-coded json string for json data validation. Each test use its own json string
0524e41 [qqsun8819] Merge remote-tracking branch 'upstream/master' into json-protocol
d203d5c [qqsun8819] [SPARK-1038] Add more fields in JsonProtocol and add tests that verify the JSON itself
[SPARK-1060] startJettyServer should explicitly use IP information
https://spark-project.atlassian.net/browse/SPARK-1060
In the current implementation, the webserver in Master/Worker is started with
val (srv, bPort) = JettyUtils.startJettyServer("0.0.0.0", port, handlers)
inside startJettyServer:
val server = new Server(currentPort) //here, the Server will take "0.0.0.0" as the hostname, i.e. will always bind to the IP address of the first NIC
this can cause wrong IP binding, e.g. if the host has two NICs, N1 and N2, the user specify the SPARK_LOCAL_IP as the N2's IP address, however, when starting the web server, for the reason stated above, it will always bind to the N1's address
Author: CodingCat <zhunansjtu@gmail.com>
== Merge branch commits ==
commit 6c6d9a8ccc9ec4590678a3b34cb03df19092029d
Author: CodingCat <zhunansjtu@gmail.com>
Date: Thu Feb 6 14:53:34 2014 -0500
startJettyServer should explicitly use IP information
Inform DAG scheduler about all started/finished tasks.
Previously, the DAG scheduler was not always informed
when tasks started and finished. The simplest example here
is for speculated tasks: the DAGScheduler was only told about
the first attempt of a task, meaning that SparkListeners were
also not told about multiple task attempts, so users can't see
what's going on with speculation in the UI. The DAGScheduler
also wasn't always told about finished tasks, so in the UI, some
tasks will never be shown as finished (this occurs, for example,
if a task set gets killed).
The other problem is that the fairness accounting was wrong
-- the number of running tasks in a pool was decreased when a
task set was considered done, even if all of its tasks hadn't
yet finished.
Author: Kay Ousterhout <kayousterhout@gmail.com>
== Merge branch commits ==
commit c8d547d0f7a17f5a193bef05f5872b9f475675c5
Author: Kay Ousterhout <kayousterhout@gmail.com>
Date: Wed Jan 15 16:47:33 2014 -0800
Addressed Reynold's review comments.
Always use a TaskEndReason (remove the option), and explicitly
signal when we don't know the reason. Also, always tell
DAGScheduler (and associated listeners) about started tasks, even
when they're speculated.
commit 3fee1e2e3c06b975ff7f95d595448f38cce97a04
Author: Kay Ousterhout <kayousterhout@gmail.com>
Date: Wed Jan 8 22:58:13 2014 -0800
Fixed broken test and improved logging
commit ff12fcaa2567c5d02b75a1d5db35687225bcd46f
Author: Kay Ousterhout <kayousterhout@gmail.com>
Date: Sun Dec 29 21:08:20 2013 -0800
Inform DAG scheduler about all finished tasks.
Previously, the DAG scheduler was not always informed
when tasks finished. For example, when a task set was
aborted, the DAG scheduler was never told when the tasks
in that task set finished. The DAG scheduler was also
never told about the completion of speculated tasks.
This led to confusion with SparkListeners because information
about the completion of those tasks was never passed on to
the listeners (so in the UI, for example, some tasks will never
be shown as finished).
The other problem is that the fairness accounting was wrong
-- the number of running tasks in a pool was decreased when a
task set was considered done, even if all of its tasks hadn't
yet finished.
Fixed warnings in test compilation.
This commit fixes two problems: a redundant import, and a
deprecated function.
Author: Kay Ousterhout <kayousterhout@gmail.com>
== Merge branch commits ==
commit da9d2e13ee4102bc58888df0559c65cb26232a82
Author: Kay Ousterhout <kayousterhout@gmail.com>
Date: Wed Feb 5 11:41:51 2014 -0800
Fixed warnings in test compilation.
This commit fixes two problems: a redundant import, and a
deprecated function.
Refactor RDD sampling and add randomSplit to RDD (update)
Replace SampledRDD by PartitionwiseSampledRDD, which accepts a RandomSampler instance as input. The current sample with/without replacement can be easily integrated via BernoulliSampler and PoissonSampler. The benefits are:
1) RDD.randomSplit is implemented in the same way, related to https://github.com/apache/incubator-spark/pull/513
2) Stratified sampling and importance sampling can be implemented in the same manner as well.
Unit tests are included for samplers and RDD.randomSplit.
This should performance better than my previous request where the BernoulliSampler creates many Iterator instances:
https://github.com/apache/incubator-spark/pull/513
Author: Xiangrui Meng <meng@databricks.com>
== Merge branch commits ==
commit e8ce957e5f0a600f2dec057924f4a2ca6adba373
Author: Xiangrui Meng <meng@databricks.com>
Date: Mon Feb 3 12:21:08 2014 -0800
more docs to PartitionwiseSampledRDD
commit fbb4586d0478ff638b24bce95f75ff06f713d43b
Author: Xiangrui Meng <meng@databricks.com>
Date: Mon Feb 3 00:44:23 2014 -0800
move XORShiftRandom to util.random and use it in BernoulliSampler
commit 987456b0ee8612fd4f73cb8c40967112dc3c4c2d
Author: Xiangrui Meng <meng@databricks.com>
Date: Sat Feb 1 11:06:59 2014 -0800
relax assertions in SortingSuite because the RangePartitioner has large variance in this case
commit 3690aae416b2dc9b2f9ba32efa465ba7948477f4
Author: Xiangrui Meng <meng@databricks.com>
Date: Sat Feb 1 09:56:28 2014 -0800
test split ratio of RDD.randomSplit
commit 8a410bc933a60c4d63852606f8bbc812e416d6ae
Author: Xiangrui Meng <meng@databricks.com>
Date: Sat Feb 1 09:25:22 2014 -0800
add a test to ensure seed distribution and minor style update
commit ce7e866f674c30ab48a9ceb09da846d5362ab4b6
Author: Xiangrui Meng <meng@databricks.com>
Date: Fri Jan 31 18:06:22 2014 -0800
minor style change
commit 750912b4d77596ed807d361347bd2b7e3b9b7a74
Author: Xiangrui Meng <meng@databricks.com>
Date: Fri Jan 31 18:04:54 2014 -0800
fix some long lines
commit c446a25c38d81db02821f7f194b0ce5ab4ed7ff5
Author: Xiangrui Meng <meng@databricks.com>
Date: Fri Jan 31 17:59:59 2014 -0800
add complement to BernoulliSampler and minor style changes
commit dbe2bc2bd888a7bdccb127ee6595840274499403
Author: Xiangrui Meng <meng@databricks.com>
Date: Fri Jan 31 17:45:08 2014 -0800
switch to partition-wise sampling for better performance
commit a1fca5232308feb369339eac67864c787455bb23
Merge: ac712e4 cf6128f
Author: Xiangrui Meng <meng@databricks.com>
Date: Fri Jan 31 16:33:09 2014 -0800
Merge branch 'sample' of github.com:mengxr/incubator-spark into sample
commit cf6128fb672e8c589615adbd3eaa3cbdb72bd461
Author: Xiangrui Meng <meng@databricks.com>
Date: Sun Jan 26 14:40:07 2014 -0800
set SampledRDD deprecated in 1.0
commit f430f847c3df91a3894687c513f23f823f77c255
Author: Xiangrui Meng <meng@databricks.com>
Date: Sun Jan 26 14:38:59 2014 -0800
update code style
commit a8b5e2021a9204e318c80a44d00c5c495f1befb6
Author: Xiangrui Meng <meng@databricks.com>
Date: Sun Jan 26 12:56:27 2014 -0800
move package random to util.random
commit ab0fa2c4965033737a9e3a9bf0a59cbb0df6a6f5
Author: Xiangrui Meng <meng@databricks.com>
Date: Sun Jan 26 12:50:35 2014 -0800
add Apache headers and update code style
commit 985609fe1a55655ad11966e05a93c18c138a403d
Author: Xiangrui Meng <meng@databricks.com>
Date: Sun Jan 26 11:49:25 2014 -0800
add new lines
commit b21bddf29850a2c006a868869b8f91960a029322
Author: Xiangrui Meng <meng@databricks.com>
Date: Sun Jan 26 11:46:35 2014 -0800
move samplers to random.IndependentRandomSampler and add tests
commit c02dacb4a941618e434cefc129c002915db08be6
Author: Xiangrui Meng <meng@databricks.com>
Date: Sat Jan 25 15:20:24 2014 -0800
add RandomSampler
commit 8ff7ba3c5cf1fc338c29ae8b5fa06c222640e89c
Author: Xiangrui Meng <meng@databricks.com>
Date: Fri Jan 24 13:23:22 2014 -0800
init impl of IndependentlySampledRDD
This fixes SPARK-1043, a bug introduced in 0.9.0
where PySpark couldn't serialize strings > 64kB.
This fix was written by @tyro89 and @bouk in #512.
This commit squashes and rebases their pull request
in order to fix some merge conflicts.
Replace the check for None Option with isDefined and isEmpty in Scala code
Propose to replace the Scala check for Option "!= None" with Option.isDefined and "=== None" with Option.isEmpty.
I think this, using method call if possible then operator function plus argument, will make the Scala code easier to read and understand.
Pass compile and tests.
This fixes an issue where collectAsMap() could
fail when called on a JavaPairRDD that was derived
by transforming a non-JavaPairRDD.
The root problem was that we were creating the
JavaPairRDD's ClassTag by casting a
ClassTag[AnyRef] to a ClassTag[Tuple2[K2, V2]].
To fix this, I cast a ClassTag[Tuple2[_, _]]
instead, since this actually produces a ClassTag
of the appropriate type because ClassTags don't
capture type parameters:
scala> implicitly[ClassTag[Tuple2[_, _]]] == implicitly[ClassTag[Tuple2[Int, Int]]]
res8: Boolean = true
scala> implicitly[ClassTag[AnyRef]].asInstanceOf[ClassTag[Tuple2[Int, Int]]] == implicitly[ClassTag[Tuple2[Int, Int]]]
res9: Boolean = false
Remove Typesafe Config usage and conf files to fix nested property names
With Typesafe Config we had the subtle problem of no longer allowing
nested property names, which are used for a few of our properties:
http://apache-spark-developers-list.1001551.n3.nabble.com/Config-properties-broken-in-master-td208.html
This PR is for branch 0.9 but should be added into master too.
(cherry picked from commit 34e911ce9a)
Signed-off-by: Patrick Wendell <pwendell@gmail.com>
Prior to this commit, if a task crashes the JVM, the task (and
all other tasks running on that executor) is marked at KILLED rather
than FAILED. As a result, the TaskSetManager will retry the task
indefiniteily rather than failing the job after maxFailures. This
commit fixes that problem by marking tasks as FAILED rather than
killed when an executor is lost.
The downside of this commit is that if task A fails because another
task running on the same executor caused the VM to crash, the failure
will incorrectly be counted as a failure of task A. This should not
be an issue because we typically set maxFailures to 3, and it is
unlikely that a task will be co-located with a JVM-crashing task
multiple times.
Remove now un-needed hostPort option
I noticed this was logging some scary error messages in various places. After I looked into it, this is no longer really used. I removed the option and re-wrote the one remaining use case (it was unnecessary there anyways).
External Sorting for Aggregator and CoGroupedRDDs (Revisited)
(This pull request is re-opened from https://github.com/apache/incubator-spark/pull/303, which was closed because Jenkins / github was misbehaving)
The target issue for this patch is the out-of-memory exceptions triggered by aggregate operations such as reduce, groupBy, join, and cogroup. The existing AppendOnlyMap used by these operations resides purely in memory, and grows with the size of the input data until the amount of allocated memory is exceeded. Under large workloads, this problem is aggravated by the fact that OOM frequently occurs only after a very long (> 1 hour) map phase, in which case the entire job must be restarted.
The solution is to spill the contents of this map to disk once a certain memory threshold is exceeded. This functionality is provided by ExternalAppendOnlyMap, which additionally sorts this buffer before writing it out to disk, and later merges these buffers back in sorted order.
Under normal circumstances in which OOM is not triggered, ExternalAppendOnlyMap is simply a wrapper around AppendOnlyMap and incurs little overhead. Only when the memory usage is expected to exceed the given threshold does ExternalAppendOnlyMap spill to disk.
SPARK-998: Support Launching Driver Inside of Standalone Mode
[NOTE: I need to bring the tests up to date with new changes, so for now they will fail]
This patch provides support for launching driver programs inside of a standalone cluster manager. It also supports monitoring and re-launching of driver programs which is useful for long running, recoverable applications such as Spark Streaming jobs. For those jobs, this patch allows a deployment mode which is resilient to the failure of any worker node, failure of a master node (provided a multi-master setup), and even failures of the applicaiton itself, provided they are recoverable on a restart. Driver information, such as the status and logs from a driver, is displayed in the UI
There are a few small TODO's here, but the code is generally feature-complete. They are:
- Bring tests up to date and add test coverage
- Restarting on failure should be optional and maybe off by default.
- See if we can re-use akka connections to facilitate clients behind a firewall
A sensible place to start for review would be to look at the `DriverClient` class which presents users the ability to launch their driver program. I've also added an example program (`DriverSubmissionTest`) that allows you to test this locally and play around with killing workers, etc. Most of the code is devoted to persisting driver state in the cluster manger, exposing it in the UI, and dealing correctly with various types of failures.
Instructions to test locally:
- `sbt/sbt assembly/assembly examples/assembly`
- start a local version of the standalone cluster manager
```
./spark-class org.apache.spark.deploy.client.DriverClient \
-j -Dspark.test.property=something \
-e SPARK_TEST_KEY=SOMEVALUE \
launch spark://10.99.1.14:7077 \
../path-to-examples-assembly-jar \
org.apache.spark.examples.DriverSubmissionTest 1000 some extra options --some-option-here -X 13
```
- Go in the UI and make sure it started correctly, look at the output etc
- Kill workers, the driver program, masters, etc.
Get rid of `Either[ActorRef, ActorSelection]'
In this pull request, instead of returning an `Either[ActorRef, ActorSelection]`, `registerOrLookup` identifies the remote actor blockingly to obtain an `ActorRef`, or throws an exception if the remote actor doesn't exist or the lookup times out (configured by `spark.akka.lookupTimeout`). This function is only called when an `SparkEnv` is constructed (instantiating driver or executor), so the blocking call is considered acceptable. Executor side `ActorSelection`s/`ActorRef`s to driver side `MapOutputTrackerMasterActor` and `BlockManagerMasterActor` are affected by this pull request.
`ActorSelection` is dangerous and should be used with care. It's only absolutely safe to send messages via an `ActorSelection` when the remote actor is stateless, so that actor incarnation is irrelevant. But as pointed by @ScrapCodes in the comments below, executor exits immediately once the connection to the driver lost, `ActorSelection`s are not harmful in this scenario. So this pull request is mostly a code style patch.
Add way to limit default # of cores used by apps in standalone mode
Also documents the spark.deploy.spreadOut option, and fixes a config option that had a dash in its name.
Although we can send messages via an ActorSelection, it would be better to identify the actor and obtain an ActorRef first, so that we can get informed earlier if the remote actor doesn't exist, and get rid of the annoying Either wrapper.
Further, divide this threshold by the number of tasks running concurrently.
Note that this does not guard against the following scenario: a new task
quickly fills up its share of the memory before old tasks finish spilling
their contents, in which case the total memory used by such maps may exceed
what was specified. Currently, spark.shuffle.safetyFraction mitigates the
effect of this.
Improvements to DStream window ops and refactoring of Spark's CheckpointSuite
- Added a new RDD - PartitionerAwareUnionRDD. Using this RDD, one can take multiple RDDs partitioned by the same partitioner and unify them into a single RDD while preserving the partitioner. So m RDDs with p partitions each will be unified to a single RDD with p partitions and the same partitioner. The preferred location for each partition of the unified RDD will be the most common preferred location of the corresponding partitions of the parent RDDs. For example, location of partition 0 of the unified RDD will be where most of partition 0 of the parent RDDs are located.
- Improved the performance of DStream's reduceByKeyAndWindow and groupByKeyAndWindow. Both these operations work by doing per-batch reduceByKey/groupByKey and then using PartitionerAwareUnionRDD to union the RDDs across the window. This eliminates a shuffle related to the window operation, which can reduce batch processing time by 30-40% for simple workloads.
- Fixed bugs and simplified Spark's CheckpointSuite. Some of the tests were incorrect and unreliable. Added missing tests for ZippedRDD. I can go into greater detail if necessary.
- Added mapSideCombine option to combineByKeyAndWindow.
Also replaced SparkConf.getOrElse with just a "get" that takes a default
value, and added getInt, getLong, etc to make code that uses this
simpler later on.
Approximate distinct count
Added countApproxDistinct() to RDD and countApproxDistinctByKey() to PairRDDFunctions to approximately count distinct number of elements and distinct number of values per key, respectively. Both functions use HyperLogLog from stream-lib for counting. Both functions take a parameter that controls the trade-off between accuracy and memory consumption. Also added Scala docs and test suites for both methods.
Bug fixes for file input stream and checkpointing
- Fixed bugs in the file input stream that led the stream to fail due to transient HDFS errors (listing files when a background thread it deleting fails caused errors, etc.)
- Updated Spark's CheckpointRDD and Streaming's CheckpointWriter to use SparkContext.hadoopConfiguration, to allow checkpoints to be written to any HDFS compatible store requiring special configuration.
- Changed the API of SparkContext.setCheckpointDir() - eliminated the unnecessary 'useExisting' parameter. Now SparkContext will always create a unique subdirectory within the user specified checkpoint directory. This is to ensure that previous checkpoint files are not accidentally overwritten.
- Fixed bug where setting checkpoint directory as a relative local path caused the checkpointing to fail.
- Got rid of global SparkContext.globalConf
- Pass SparkConf to serializers and compression codecs
- Made SparkConf public instead of private[spark]
- Improved API of SparkContext and SparkConf
- Switched executor environment vars to be passed through SparkConf
- Fixed some places that were still using system properties
- Fixed some tests, though others are still failing
This still fails several tests in core, repl and streaming, likely due
to properties not being set or cleared correctly (some of the tests run
fine in isolation).
The rest of the SparkListener events are named with "SparkListener"
as the prefix of the name; this commit renames the StageCompleted
event to SparkListenerStageCompleted for consistency.
Deduplicate Local and Cluster schedulers.
The code in LocalScheduler/LocalTaskSetManager was nearly identical
to the code in ClusterScheduler/ClusterTaskSetManager. The redundancy
made making updating the schedulers unnecessarily painful and error-
prone. This commit combines the two into a single TaskScheduler/
TaskSetManager.
Unfortunately the diff makes this change look much more invasive than it is -- TaskScheduler.scala is only superficially changed (names updated, overrides removed) from the old ClusterScheduler.scala, and the same with
TaskSetManager.scala.
Thanks @rxin for suggesting this change!
Track and report task result serialisation time.
- DirectTaskResult now has a ByteBuffer valueBytes instead of a T value.
- DirectTaskResult now has a member function T value() that deserialises valueBytes.
- Executor serialises value into a ByteBuffer and passes it to DTR's ctor.
- Executor tracks the time taken to do so and puts it in a new field in TaskMetrics.
- StagePage now reports serialisation time from TaskMetrics along with the other things it reported.
Change the implementation to use runJob instead of PartitionPruningRDD.
Also update the unit tests and the python take implementation
to use the new interface.
Fix for spark.task.maxFailures not enforced correctly.
Docs at http://spark.incubator.apache.org/docs/latest/configuration.html say:
```
spark.task.maxFailures
Number of individual task failures before giving up on the job. Should be greater than or equal to 1. Number of allowed retries = this value - 1.
```
Previous implementation worked incorrectly. When for example `spark.task.maxFailures` was set to 1, the job was aborted only after the second task failure, not after the first one.
- Made file stream more robust to transient failures.
- Changed Spark.setCheckpointDir API to not have the second
'useExisting' parameter. Spark will always create a unique directory
for checkpointing underneath the directory provide to the funtion.
- Fixed bug wrt local relative paths as checkpoint directory.
- Made DStream and RDD checkpointing use
SparkContext.hadoopConfiguration, so that more HDFS compatible
filesystems are supported for checkpointing.
stageId <--> jobId mapping in DAGScheduler
Okay, I think this one is ready to go -- or at least it's ready for review and discussion. It's a carry-over of https://github.com/mesos/spark/pull/842 with updates for the newer job cancellation functionality. The prior discussion still applies. I've actually changed the job cancellation flow a bit: Instead of ``cancelTasks`` going to the TaskScheduler and then ``taskSetFailed`` coming back to the DAGScheduler (resulting in ``abortStage`` there), the DAGScheduler now takes care of figuring out which stages should be cancelled, tells the TaskScheduler to cancel tasks for those stages, then does the cleanup within the DAGScheduler directly without the need for any further prompting by the TaskScheduler.
I know of three outstanding issues, each of which can and should, I believe, be handled in follow-up pull requests:
1) https://spark-project.atlassian.net/browse/SPARK-960
2) JobLogger should be re-factored to eliminate duplication
3) Related to 2), the WebUI should also become a consumer of the DAGScheduler's new understanding of the relationship between jobs and stages so that it can display progress indication and the like grouped by job. Right now, some of this information is just being sent out as part of ``SparkListenerJobStart`` messages, but more or different job <--> stage information may need to be exported from the DAGScheduler to meet listeners needs.
Except for the eventQueue -> Actor commit, the rest can be cherry-picked almost cleanly into branch-0.8. A little merging is needed in MapOutputTracker and the DAGScheduler. Merged versions of those files are in aba2b40ce0
Note that between the recent Actor change in the DAGScheduler and the cleaning up of DAGScheduler data structures on job completion in this PR, some races have been introduced into the DAGSchedulerSuite. Those tests usually pass, and I don't think that better-behaved code that doesn't directly inspect DAGScheduler data structures should be seeing any problems, but I'll work on fixing DAGSchedulerSuite as either an addition to this PR or as a separate request.
UPDATE: Fixed the race that I introduced. Created a JIRA issue (SPARK-965) for the one that was introduced with the switch to eventProcessorActor in the DAGScheduler.
...and make sure that DAGScheduler data structures are cleaned up on job completion.
Initial effort and discussion at https://github.com/mesos/spark/pull/842
Re-enable zk:// urls for Mesos SparkContexts
This was broken in PR #71 when we explicitly disallow anything that didn't fit a mesos:// url.
Although it is not really clear that a zk:// url should match Mesos, it is what the docs say and it is necessary for backwards compatibility.
Additionally added a unit test for the creation of all types of TaskSchedulers. Since YARN and Mesos are not necessarily available in the system, they are allowed to pass as long as the YARN/Mesos code paths are exercised.
OpenHashSet fixes
Incorporated ideas from pull request #200.
- Use Murmur Hash 3 finalization step to scramble the bits of HashCode
instead of the simpler version in java.util.HashMap; the latter one
had trouble with ranges of consecutive integers. Murmur Hash 3 is used
by fastutil.
- Don't check keys for equality when re-inserting due to growing the
table; the keys will already be unique.
- Remember the grow threshold instead of recomputing it on each insert
Also added unit tests for size estimation for specialized hash sets and maps.
Use the proper partition index in mapPartitionsWIthIndex
mapPartitionsWithIndex uses TaskContext.partitionId as the partition index. TaskContext.partitionId used to be identical to the partition index in a RDD. However, pull request #186 introduced a scenario (with partition pruning) that the two can be different. This pull request uses the right partition index in all mapPartitionsWithIndex related calls.
Also removed the extra MapPartitionsWIthContextRDD and put all the mapPartitions related functionality in MapPartitionsRDD.
XORShift RNG with unit tests and benchmark
This patch was introduced to address SPARK-950 - the discussion below the ticket explains not only the rationale, but also the design and testing decisions: https://spark-project.atlassian.net/browse/SPARK-950
To run unit test, start SBT console and type:
compile
test-only org.apache.spark.util.XORShiftRandomSuite
To run benchmark, type:
project core
console
Once the Scala console starts, type:
org.apache.spark.util.XORShiftRandom.benchmark(100000000)
XORShiftRandom is also an object with a main method taking the
number of iterations as an argument, so you can also run it
from the command line.
Also changed the semantics of the index parameter in mapPartitionsWithIndex from the partition index of the output partition to the partition index in the current RDD.
PartitionPruningRDD is using index from parent
I was getting a ArrayIndexOutOfBoundsException exception after doing union on pruned RDD. The index it was using on the partition was the index in the original RDD not the new pruned RDD.
To run unit test, start SBT console and type:
compile
test-only org.apache.spark.util.XORShiftRandomSuite
To run benchmark, type:
project core
console
Once the Scala console starts, type:
org.apache.spark.util.XORShiftRandom.benchmark(100000000)
Migrate the daemon thread started by DAGScheduler to Akka actor
`DAGScheduler` adopts an event queue and a daemon thread polling the it to process events sent to a `DAGScheduler`. This is a classical actor use case. By migrating this thread to Akka actor, we may benefit from both cleaner code and better performance (context switching cost of Akka actor is much less than that of a native thread).
But things become a little complicated when taking existing test code into consideration.
Code in `DAGSchedulerSuite` is somewhat tightly coupled with `DAGScheduler`, and directly calls `DAGScheduler.processEvent` instead of posting event messages to `DAGScheduler`. To minimize code change, I chose to let the actor to delegate messages to `processEvent`. Maybe this doesn't follow conventional actor usage, but I tried to make it apparently correct.
Another tricky part is that, since `DAGScheduler` depends on the `ActorSystem` provided by its field `env`, `env` cannot be null. But the `dagScheduler` field created in `DAGSchedulerSuite.before` was given a null `env`. What's more, `BlockManager.blockIdsToBlockManagers` checks whether `env` is null to determine whether to run the production code or the test code (bad smell here, huh?). I went through all callers of `BlockManager.blockIdsToBlockManagers`, and made sure that if `env != null` holds, then `blockManagerMaster == null` must also hold. That's the logic behind `BlockManager.scala` [line 896](https://github.com/liancheng/incubator-spark/compare/dagscheduler-actor-refine?expand=1#diff-2b643ea78c1add0381754b1f47eec132L896).
At last, since `DAGScheduler` instances are always `start()`ed after creation, I removed the `start()` method, and starts the `eventProcessActor` within the constructor.
Include appId in executor cmd line args
add the appId back into the executor cmd line args.
I also made a pretty lame regression test, just to make sure it doesn't get dropped in the future. not sure it will run on the build server, though, b/c `ExecutorRunner.buildCommandSeq()` expects to be abel to run the scripts in `bin`.
add javadoc to JobLogger, and some small fix
against Spark-941
add javadoc to JobLogger, output more info for RDD, modify recordStageDepGraph to avoid output duplicate stage dependency information
(cherry picked from commit 518cf22eb2)
Signed-off-by: Reynold Xin <rxin@apache.org>
For some reason, even calling
java.nio.Files.createTempDirectory().getFile.deleteOnExit()
does not delete the directory on exit. Guava's analagous function
seems to work, however.
Overhead of each shuffle block for consolidation has been reduced from >300 bytes
to 8 bytes (1 primitive Long). Verified via profiler testing with 1 mil shuffle blocks,
net overhead was ~8,400,000 bytes.
Despite the memory-optimized implementation incurring extra CPU overhead, the runtime
of the shuffle phase in this test was only around 2% slower, while the reduce phase
was 40% faster, when compared to not using any shuffle file consolidation.
The code in LocalScheduler/LocalTaskSetManager was nearly identical
to the code in ClusterScheduler/ClusterTaskSetManager. The redundancy
made making updating the schedulers unnecessarily painful and error-
prone. This commit combines the two into a single TaskScheduler/
TaskSetManager.
This patch adds an operator called repartition with more straightforward
semantics than the current `coalesce` operator. There are a few use cases
where this operator is useful:
1. If a user wants to increase the number of partitions in the RDD. This
is more common now with streaming. E.g. a user is ingesting data on one
node but they want to add more partitions to ensure parallelism of
subsequent operations across threads or the cluster.
Right now they have to call rdd.coalesce(numSplits, shuffle=true) - that's
super confusing.
2. If a user has input data where the number of partitions is not known. E.g.
> sc.textFile("some file").coalesce(50)....
This is both vague semantically (am I growing or shrinking this RDD) but also,
may not work correctly if the base RDD has fewer than 50 partitions.
The new operator forces shuffles every time, so it will always produce exactly
the number of new partitions. It also throws an exception rather than silently
not-working if a bad input is passed.
I am currently adding streaming tests (requires refactoring some of the test
suite to allow testing at partition granularity), so this is not ready for
merge yet. But feedback is welcome.
This patch fixes a bug where the Spark UI didn't display the correct number of total
tasks if the number of tasks in a Stage doesn't equal the number of RDD partitions.
It also cleans up the listener API a bit by embedding this information in the
StageInfo class rather than passing it seperately.
This commit adds a set of calls using the SparkListener interface
that indicate when a task is remotely fetching results, so that
we can display this (potentially time-consuming) phase of execution
to users through the UI.
Previously, MapOutputTracker contained fields and methods that
were only applicable to the master or worker instances. This
commit introduces a MasterMapOutputTracker class to prevent
the master-specific methods from being accessed on workers.
I also renamed a few methods and made others protected/private.
Job cancellation via job group id.
This PR adds a simple API to group together a set of jobs belonging to a thread and threads spawned from it. It also allows the cancellation of all jobs in this group.
An example:
sc.setJobDescription("this_is_the_group_id", "some job description")
sc.parallelize(1 to 10000, 2).map { i => Thread.sleep(10); i }.count()
In a separate thread:
sc.cancelJobGroup("this_is_the_group_id")
This is an unfortunately invasive change which converts all of our BlockId
strings into actual BlockId types. Here are some advantages of doing this now:
+ Type safety
+ Code clarity - it's now obvious what the key of a shuffle or rdd block is,
for instance. Additionally, appearing in tuple/map type signatures is a big
readability bonus. A Seq[(String, BlockStatus)] is not very clear.
Further, we can now use more Scala features, like matching on BlockId types.
+ Explicit usage - we can now formally tell where various BlockIds are being used
(without doing string searches); this makes updating current BlockIds a much
clearer process, and compiler-supported.
(I'm looking at you, shuffle file consolidation.)
+ It will only get harder to make this change as time goes on.
Since this touches a lot of files, it'd be best to either get this patch
in quickly or throw it on the ground to avoid too many secondary merge conflicts.
Standalone Scheduler fault tolerance using ZooKeeper
This patch implements full distributed fault tolerance for standalone scheduler Masters.
There is only one master Leader at a time, which is actively serving scheduling
requests. If this Leader crashes, another master will eventually be elected, reconstruct
the state from the first Master, and continue serving scheduling requests.
Leader election is performed using the ZooKeeper leader election pattern. We try to minimize
the use of ZooKeeper and the assumptions about ZooKeeper's behavior, so there is a layer of
retries and session monitoring on top of the ZooKeeper client.
Master failover follows directly from the single-node Master recovery via the file
system (patch d5a96fe), save that the Master state is stored in ZooKeeper instead.
Configuration:
By default, no recovery mechanism is enabled (spark.deploy.recoveryMode = NONE).
By setting spark.deploy.recoveryMode to ZOOKEEPER and setting spark.deploy.zookeeper.url
to an appropriate ZooKeeper URL, ZooKeeper recovery mode is enabled.
By setting spark.deploy.recoveryMode to FILESYSTEM and setting spark.deploy.recoveryDirectory
to an appropriate directory accessible by the Master, we will keep the behavior of from d5a96fe.
Additionally, places where a Master could be specificied by a spark:// url can now take
comma-delimited lists to specify backup masters. Note that this is only used for registration
of NEW Workers and application Clients. Once a Worker or Client has registered with the
Master Leader, it is "in the system" and will never need to register again.
Conflicts:
bagel/pom.xml
core/pom.xml
core/src/test/scala/org/apache/spark/ui/UISuite.scala
examples/pom.xml
mllib/pom.xml
pom.xml
project/SparkBuild.scala
repl/pom.xml
streaming/pom.xml
tools/pom.xml
In scala 2.10, a shorter representation is used for naming artifacts
so changed to shorter scala version for artifacts and made it a property in pom.
Resolving package conflicts with hadoop 0.23.9
Hadoop 0.23.9 is having a package conflict with easymock's dependencies.
(cherry picked from commit 023e3fdf00)
Signed-off-by: Reynold Xin <rxin@apache.org>
One major change was the use of messages instead of raw functions as the
parameter of Akka scheduled timers. Since messages are serialized, unlike
raw functions, the behavior is easier to think about and doesn't cause
race conditions when exceptions are thrown.
Another change is to avoid using global pointers that might change without
a lock.
Currently PythonPartitioner determines partition ID by hashing a
byte-array representation of PySpark's key. This PR lets
PythonPartitioner use the actual partition ID, which is required e.g.
for sorting via PySpark.
Improved organization of scheduling packages.
This commit does not change any code -- only file organization.
Please let me know if there was some masterminded strategy behind
the existing organization that I failed to understand!
There are two components of this change:
(1) Moving files out of the cluster package, and down
a level to the scheduling package. These files are all used by
the local scheduler in addition to the cluster scheduler(s), so
should not be in the cluster package. As a result of this change,
none of the files in the local package reference files in the
cluster package.
(2) Moving the mesos package to within the cluster package.
The mesos scheduling code is for a cluster, and represents a
specific case of cluster scheduling (the Mesos-related classes
often subclass cluster scheduling classes). Thus, the most logical
place for it seems to be within the cluster package.
The one thing about the scheduling code that seems a little funny to me
is the naming of the SchedulerBackends. The StandaloneSchedulerBackend
is not just for Standalone mode, but instead is used by Mesos coarse grained
mode and Yarn, and the backend that *is* just for Standalone mode is instead called SparkDeploySchedulerBackend. I didn't change this because I wasn't sure if there
was a reason for this naming that I'm just not aware of.
This commit does not change any code -- only file organization.
There are two components of this change:
(1) Moving files out of the cluster package, and down
a level to the scheduling package. These files are all used by
the local scheduler in addition to the cluster scheduler(s), so
should not be in the cluster package. As a result of this change,
none of the files in the local package reference files in the
cluster package.
(2) Moving the mesos package to within the cluster package.
The mesos scheduling code is for a cluster, and represents a
specific case of cluster scheduling (the Mesos-related classes
often subclass cluster scheduling classes). Thus, the most logical
place for it is within the cluster package.
This change requires adding an extra failure mode: tasks can complete
successfully, but the result gets lost or flushed from the block manager
before it's been fetched.
In MapOutputTrackerSuite, the "remote fetch" test sets spark.driver.port
and spark.hostPort, assuming that they will be cleared by
LocalSparkContext. However, the test never sets sc, so it remains null,
causing LocalSparkContext to skip clearing these properties. Subsequent
tests therefore fail with java.net.BindException: "Address already in
use".
This commit makes LocalSparkContext clear the properties even if sc is
null.
Include the useful tip that if shuffle=true, coalesce can actually
increase the number of partitions.
This makes coalesce more like a generic `RDD.repartition` operation.
(Ideally this `RDD.repartition` could automatically choose either a coalesce or
a shuffle if numPartitions was either less than or greater than, respectively,
the current number of partitions.)
Caching the results of local actions (e.g., rdd.first()) causes the driver to
store entire partitions in its own memory, which may be highly constrained.
This patch simply makes the CacheManager avoid caching the result of all locally-run computations.
This commit makes Spark invocation saner by using an assembly JAR to
find all of Spark's dependencies instead of adding all the JARs in
lib_managed. It also packages the examples into an assembly and uses
that as SPARK_EXAMPLES_JAR. Finally, it replaces the old "run" script
with two better-named scripts: "run-examples" for examples, and
"spark-class" for Spark internal classes (e.g. REPL, master, etc). This
is also designed to minimize the confusion people have in trying to use
"run" to run their own classes; it's not meant to do that, but now at
least if they look at it, they can modify run-examples to do a decent
job for them.
As part of this, Bagel's examples are also now properly moved to the
examples package instead of bagel.
- Use SPARK_PUBLIC_DNS environment variable if set (for EC2)
- Use a non-ephemeral port (3030 instead of 33000) by default
- Updated test to use non-ephemeral port too
- When a resourceOffers() call has multiple offers, force the TaskSets
to consider them in increasing order of locality levels so that they
get a chance to launch stuff locally across all offers
- Simplify ClusterScheduler.prioritizeContainers
- Add docs on the new configuration options
- Replace use of hostPort vs host in Task.preferredLocations with a
TaskLocation class that contains either an executorId and a host or
just a host. This is part of a bigger effort to eliminate hostPort
based data structures and just use executorID, since the hostPort vs
host stuff is confusing (and not checkable with static typing, leading
to ugly debug code), and hostPorts are not provided by Mesos.
- Replaced most hostPort-based data structures and fields as above.
- Simplified ClusterTaskSetManager to deal with preferred locations in a
more concise way and generally be more concise.
- Updated the way ClusterTaskSetManager handles racks: instead of
enqueueing a task to a separate queue for all the hosts in the rack,
which would create lots of large queues, have one queue per rack name.
- Removed non-local fallback stuff in ClusterScheduler that tried to
launch less-local tasks on a node once the local ones were all
assigned. This change didn't work because many cluster schedulers send
offers for just one node at a time (even the standalone and YARN ones
do so as nodes join the cluster one by one). Thus, lots of non-local
tasks would be assigned even though a node with locality for them
would be able to receive tasks just a short time later.
- Renamed MapOutputTracker "generations" to "epochs".
These are used all over the place now and they are not specific to memory at all.
memoryBytesToString --> bytesToString
memoryMegabytesToString --> megabytesToString
1. Renamed SparkContext.addLocalProperty to setLocalProperty. And allow this function to unset a property.
2. Renamed SparkContext.setDescription to setCurrentJobDescription.
3. Throw an exception if the fair scheduler allocation file is invalid.
1. Set akka log level to ERROR before shutting down the actorSystem.
This avoids akka log messages (like Spray) from falling back to INFO
on the Stdout logger
2. Initialize netty to use SLF4J in LocalSparkContext. This ensures that
stack trace thrown during shutdown is handled by SLF4J instead of stdout
Before, when withReplacement was set to true, we would not get a sample
bigger than the RDD's count().
Conflicts:
core/src/main/scala/spark/RDD.scala
core/src/test/scala/spark/RDDSuite.scala
- Split SPARK_JAVA_OPTS into multiple command-line arguments if it
contains spaces; this splitting follows quoting rules in bash
- Add the Scala JARs to the classpath if they're not in the CLASSPATH
variable because the ExecutorRunner is launched with "scala" (this can
happen when using local-cluster URLs in spark-shell)
The old version reused the object within each task, leading to
overwriting of the object when a mutable type is used, which is expected
to be common in fold.
Conflicts:
core/src/test/scala/spark/ShuffleSuite.scala