hadoop/hadoop-yarn-project/CHANGES.txt

651 lines
21 KiB
Plaintext

Hadoop YARN Change Log
Release 2.0.5-beta - UNRELEASED
INCOMPATIBLE CHANGES
YARN-396. Rationalize AllocateResponse in RM Scheduler API. (Zhijie Shen
via hitesh)
YARN-439. Flatten NodeHeartbeatResponse. (Xuan Gong via sseth)
YARN-440. Flatten RegisterNodeManagerResponse. (Xuan Gong via sseth)
NEW FEATURES
IMPROVEMENTS
YARN-365. Change NM heartbeat handling to not generate a scheduler event
on each heartbeat. (Xuan Gong via sseth)
YARN-380. Fix yarn node -status output to be better readable. (Omkar Vinit
Joshi via vinodkv)
YARN-410. Fixed RM UI so that the new lines diagnostics for a failed app on
the per-application page are translated to html line breaks. (Omkar Vinit
Joshi via vinodkv)
YARN-198. Added a link to RM pages from the NodeManager web app. (Jian He
via vinodkv)
YARN-237. Refreshing the RM page forgets how many rows I had in my
Datatables (jian he via bobby)
YARN-481. Add AM Host and RPC Port to ApplicationCLI Status Output
(Chris Riccomini via bikas)
YARN-297. Improve hashCode implementations for PB records. (Xuan Gong via
hitesh)
YARN-417. Create AMRMClient wrapper that provides asynchronous callbacks.
(Sandy Ryza via bikas)
YARN-497. Yarn unmanaged-am launcher jar does not define a main class in
its manifest (Hitesh Shah via bikas)
YARN-469. Make scheduling mode in FS pluggable. (kkambatl via tucu)
YARN-450. Define value for * in the scheduling protocol (Zhijie Shen via
bikas)
YARN-475. Remove a unused constant in the public API -
ApplicationConstants.AM_APP_ATTEMPT_ID_ENV. (Hitesh Shah via vinodkv)
YARN-309. Changed NodeManager to obtain heart-beat interval from the
ResourceManager. (Xuan Gong via vinodkv)
OPTIMIZATIONS
BUG FIXES
YARN-383. AMRMClientImpl should handle null rmClient in stop()
(Hitesh Shah via sseth)
YARN-385. Add missing fields - location and #containers to
ResourceRequestPBImpl's toString(). (Sandy Ryza via sseth)
YARN-377. Use the new StringUtils methods added by HADOOP-9252 and fix
TestContainersMonitor. (Chris Nauroth via szetszwo)
YARN-391. Formatting fixes for LCEResourceHandler classes.
(Steve Loughran via sseth)
YARN-390. ApplicationCLI and NodeCLI hard-coded platform-specific line
separator causes test failures on Windows. (Chris Nauroth via suresh)
YARN-406. Fix TestRackResolver to function in networks where "host1"
resolves to a valid host. (Hitesh Shah via sseth)
YARN-376. Fixes a bug which would prevent the NM knowing about completed
containers and applications. (Jason Lowe via sseth)
YARN-196. Nodemanager should be more robust in handling connection failure
to ResourceManager when a cluster is started (Xuan Gong via hitesh)
YARN-485. TestProcfsProcessTree#testProcessTree() doesn't wait long enough
for the process to die. (kkambatl via tucu)
YARN-470. Support a way to disable resource monitoring on the NodeManager.
(Siddharth Seth via hitesh)
YARN-71. Fix the NodeManager to clean up local-dirs on restart.
(Xuan Gong via sseth)
YARN-378. Fix RM to make the AM max attempts/retries to be configurable
per application by clients. (Zhijie Shen via vinodkv)
YARN-498. Unmanaged AM launcher does not set various constants in env for
an AM, also does not handle failed AMs properly. (Hitesh Shah via bikas)
YARN-496. Fair scheduler configs are refreshed inconsistently in
reinitialize. (Sandy Ryza via tomwhite)
YARN-474. Fix CapacityScheduler to trigger application-activation when
am-resource-percent configuration is refreshed. (Zhijie Shen via vinodkv)
YARN-209. Fix CapacityScheduler to trigger application-activation when
the cluster capacity changes. (Zhijie Shen via vinodkv)
YARN-24. Nodemanager fails to start if log aggregation enabled and
namenode unavailable. (sandyr via tucu)
YARN-515. Node Manager not getting the master key. (Robert Joseph Evans
via jlowe)
Release 2.0.4-alpha - UNRELEASED
INCOMPATIBLE CHANGES
NEW FEATURES
IMPROVEMENTS
OPTIMIZATIONS
BUG FIXES
YARN-429. capacity-scheduler config missing from yarn-test artifact.
(sseth via hitesh)
Release 2.0.3-alpha - 2013-02-06
INCOMPATIBLE CHANGES
NEW FEATURES
YARN-145. Add a Web UI to the fair share scheduler. (Sandy Ryza via tomwhite)
YARN-3. Add support for CPU isolation/monitoring of containers.
(adferguson via tucu)
YARN-230. RM Restart phase 1 - includes support for saving/restarting all
applications on an RM bounce. (Bikas Saha via acmurthy)
YARN-103. Add a yarn AM-RM client module. (Bikas Saha via sseth)
YARN-286. Add a YARN ApplicationClassLoader. (tomwhite)
YARN-2. Enhanced CapacityScheduler to account for CPU alongwith memory for
multi-dimensional resource scheduling. (acmurthy)
YARN-328. Use token request messages defined in hadoop common. (suresh)
YARN-231. RM Restart - Add FS-based persistent store implementation for
RMStateStore (Bikas Saha via hitesh)
IMPROVEMENTS
YARN-223. Update process tree instead of getting new process trees.
(Radim Kolar via llu)
YARN-57. Allow process-tree based resource calculation et al. to be
pluggable to support it on multiple platforms. (Radim Kolar via acmurthy)
YARN-78. Changed UnManagedAM application to use YarnClient. (Bikas Saha via
vinodkv)
YARN-93. Fixed RM to propagate diagnostics from applications that have
finished but failed (Jason Lowe via vinodkv).
YARN-28. Fixed TestCompositeService to not depend on test-order and thus
made it pass on JDK7 (Thomas Graves via vinodkv).
YARN-82. Change the default local and log dirs to be based on
hadoop.tmp.dir and yarn.log.dir. (Hemanth Yamijala via sseth)
YARN-53. Added the missing getGroups API to ResourceManager. (Bo Wang via
vinodkv)
YARN-116. Add the ability to change the RM include/exclude file without
a restart. (xieguiming and Harsh J via sseth)
YARN-23. FairScheduler: FSQueueSchedulable#updateDemand() - potential
redundant aggregation. (kkambatl via tucu)
YARN-127. Move RMAdmin tool to its correct location - the client module.
(vinodkv)
YARN-40. Provided support for missing YARN commands (Devaraj K and Vinod
Kumar Vavilapalli via vinodkv)
YARN-33. Change LocalDirsHandlerService to validate the configured local and
log dirs. (Mayank Bansal via sseth)
YARN-94. Modify DistributedShell to point to main-class by default, clean up
the help message, and hard-code the AM class. (Hitesh Shah via vinodkv)
YARN-146. Add unit tests for computing fair share in the fair scheduler.
(Sandy Ryza via tomwhite)
HADOOP-8911. CRLF characters in source and text files.
(Raja Aluri via suresh)
YARN-136. Make ClientToAMTokenSecretManager part of RMContext (Vinod Kumar
Vavilapalli via sseth)
YARN-183. Clean up fair scheduler code. (Sandy Ryza via tomwhite)
YARN-129. Simplify classpath construction for mini YARN tests. (tomwhite)
YARN-254. Update fair scheduler web UI for hierarchical queues.
(sandyr via tucu)
YARN-315. Using the common security token protobuf definition from hadoop
common. (Suresh Srinivas via vinodkv)
YARN-170. Change NodeManager stop to be reentrant. (Sandy Ryza via vinodkv)
YARN-331. Fill in missing fair scheduler documentation. (sandyr via tucu)
YARN-277. Use AMRMClient in DistributedShell to exemplify the approach.
(Bikas Saha via hitesh)
YARN-360. Allow apps to concurrently register tokens for renewal.
(Daryn Sharp via sseth)
OPTIMIZATIONS
BUG FIXES
YARN-131. Fix incorrect ACL properties in capacity scheduler documentation.
(Ahmed Radwan via sseth)
YARN-102. Move the apache header to the top of the file in MemStore.java.
(Devaraj K via sseth)
YARN-134. ClientToAMSecretManager creates keys without checking for
validity of the appID. (Vinod Kumar Vavilapalli via sseth)
YARN-30. Fixed tests verifying web-services to work on JDK7. (Thomas Graves
via vinodkv)
YARN-150. Fixes AppRejectedTransition does not unregister a rejected
app-attempt from the ApplicationMasterService (Bikas Saha via sseth)
YARN-140. Add capacity-scheduler-default.xml to provide a default set of
configurations for the capacity scheduler. (ahmed via tucu)
YARN-179. Fix some unit test failures. (Vinod Kumar Vavilapalli via sseth)
YARN-181. Fixed eclipse settings broken by capacity-scheduler.xml move via
YARN-140. (Siddharth Seth via vinodkv)
YARN-169. Update log4j.appender.EventCounter to use
org.apache.hadoop.log.metrics.EventCounter (Anthony Rojas via tomwhite)
YARN-184. Remove unnecessary locking in fair scheduler, and address
findbugs excludes. (sandyr via tucu)
YARN-224. Fair scheduler logs too many nodeUpdate INFO messages.
(Sandy Ryza via tomwhite)
YARN-222. Fair scheduler should create queue for each user by default.
(Sandy Ryza via tomwhite)
MAPREDUCE-4778. Fair scheduler event log is only written if directory
exists on HDFS. (Sandy Ryza via tomwhite)
YARN-229. Remove old unused RM recovery code. (Bikas Saha via acmurthy)
YARN-187. Add hierarchical queues to the fair scheduler.
(Sandy Ryza via tomwhite)
YARN-72. NM should handle cleaning up containers when it shuts down.
(Sandy Ryza via tomwhite)
YARN-267. Fix fair scheduler web UI. (Sandy Ryza via tomwhite)
YARN-264. y.s.rm.DelegationTokenRenewer attempts to renew token even
after removing an app. (kkambatl via tucu)
YARN-271. Fair scheduler hits IllegalStateException trying to reserve
different apps on same node. (Sandy Ryza via tomwhite)
YARN-272. Fair scheduler log messages try to print objects without
overridden toString methods. (sandyr via tucu)
YARN-278. Fair scheduler maxRunningApps config causes no apps to make
progress. (sandyr via tucu)
YARN-282. Fair scheduler web UI double counts Apps Submitted.
(sandyr via tucu)
YARN-283. Fair scheduler fails to get queue info without root prefix.
(sandyr via tucu)
YARN-192. Node update causes NPE in the fair scheduler.
(Sandy Ryza via tomwhite)
YARN-288. Fair scheduler queue doesn't accept any jobs when ACLs are
configured. (Sandy Ryza via tomwhite)
YARN-300. After YARN-271, fair scheduler can infinite loop and not
schedule any application. (Sandy Ryza via tomwhite)
YARN-301. Fair scheduler throws ConcurrentModificationException when
iterating over app's priorities. (Sandy Ryza via tomwhite)
YARN-217. Fix RMAdmin protocol description to make it work in secure mode
also. (Devaraj K via vinodkv)
YARN-253. Fixed container-launch to not fail when there are no local
resources to localize. (Tom White via vinodkv)
YARN-330. Fix flakey test: TestNodeManagerShutdown#testKillContainersOnShutdown.
(Sandy Ryza via hitesh)
YARN-335. Fair scheduler doesn't check whether rack needs containers
before assigning to node. (Sandy Ryza via tomwhite)
YARN-336. Fair scheduler FIFO scheduling within a queue only allows 1
app at a time. (Sandy Ryza via tomwhite)
YARN-135. Client tokens should be per app-attempt, and should be
unregistered on App-finish. (vinodkv via sseth)
YARN-302. Fair scheduler assignmultiple should default to false. (sandyr via tucu)
YARN-372. Move InlineDispatcher from hadoop-yarn-server-resourcemanager to
hadoop-yarn-common (sseth via hitesh)
YARN-370. Fix SchedulerUtils to correctly round up the resource for
containers. (Zhijie Shen via acmurthy)
YARN-355. Fixes a bug where RM app submission could jam under load.
(Daryn Sharp via sseth)
Release 2.0.2-alpha - 2012-09-07
YARN-9. Rename YARN_HOME to HADOOP_YARN_HOME. (vinodkv via acmurthy)
NEW FEATURES
YARN-1. Promote YARN to be a sub-project of Apache Hadoop. (acmurthy)
IMPROVEMENTS
YARN-29. Add a yarn-client module. (Vinod Kumar Vavilapalli via sseth)
YARN-10. Fix DistributedShell module to not have a dependency on
hadoop-mapreduce-client-core. (Hitesh Shah via vinodkv)
YARN-80. Add support for delaying rack-local containers in
CapacityScheduler. (acmurthy)
YARN-137. Change the default YARN scheduler to be the CapacityScheduler.
(sseth via acmurthy)
OPTIMIZATIONS
BUG FIXES
YARN-13. Fix pom versions for YARN in branch-2 (todd)
MAPREDUCE-2374. "Text File Busy" errors launching MR tasks. (Andy Isaacson
via atm)
YARN-12. Fix findbugs warnings in FairScheduler. (Junping Du via acmurthy)
YARN-22. Fix ContainerLogs to work if the log-dir is specified as a URI.
(Mayank Bansal via sseth)
YARN-37. Change TestRMAppTransitions to use the DrainDispatcher.
(Mayank Bansal via sseth)
YARN-79. Implement close on all clients to YARN so that RPC clients don't
throw exceptions on shut-down. (Vinod Kumar Vavilapalli)
YARN-42. Modify NM's non-aggregating logs' handler to stop properly so that
NMs don't get NPEs on startup errors. (Devaraj K via vinodkv)
YARN-15. Updated default classpath for YARN applications to reflect split of
YARN into a sub-project. (Arun C Murthy via vinodkv)
YARN-75. Modified ResourceManager's RMContainer to handle a valid RELEASE
event at RUNNING state. (Siddharth Seth via vinodkv)
YARN-138. Ensure default values for minimum/maximum container sizes is
sane. (harsh & sseth via acmurthy)
Release 0.23.7 - UNRELEASED
INCOMPATIBLE CHANGES
NEW FEATURES
IMPROVEMENTS
YARN-133 Update web services docs for RM clusterMetrics (Ravi Prakash via
kihwal)
YARN-249. Capacity Scheduler web page should show list of active users per
queue like it used to (in 1.x) (Ravi Prakash via tgraves)
YARN-236. RM should point tracking URL to RM web page when app fails to
start (Jason Lowe via jeagles)
YARN-269. Resource Manager not logging the health_check_script result when
taking it out (Jason Lowe via kihwal)
YARN-227. Application expiration difficult to debug for end-users
(Jason Lowe via jeagles)
YARN-443. allow OS scheduling priority of NM to be different than the
containers it launches (tgraves)
YARN-468. coverage fix for org.apache.hadoop.yarn.server.webproxy.amfilter
(Aleksey Gorshkov via bobby)
YARN-200. yarn log does not output all needed information, and is in a
binary format (Ravi Prakash via jlowe)
OPTIMIZATIONS
YARN-357. App submission should not be synchronized (daryn)
BUG FIXES
YARN-343. Capacity Scheduler maximum-capacity value -1 is invalid (Xuan
Gong via tgraves)
YARN-364. AggregatedLogDeletionService can take too long to delete logs
(jlowe)
YARN-362. Unexpected extra results when using webUI table search (Ravi
Prakash via jlowe)
YARN-400. RM can return null application resource usage report leading to
NPE in client (Jason Lowe via tgraves)
YARN-426. Failure to download a public resource prevents further downloads
(Jason Lowe via bobby)
YARN-448. Remove unnecessary hflush from log aggregation (Kihwal Lee via
bobby)
YARN-345. Many InvalidStateTransitonException errors for ApplicationImpl
in Node Manager (Robert Parker via jlowe)
YARN-109. .tmp file is not deleted for localized archives (Mayank Bansal
via bobby)
YARN-460. CS user left in list of active users for the queue even when
application finished (tgraves)
Release 0.23.6 - UNRELEASED
INCOMPATIBLE CHANGES
NEW FEATURES
IMPROVEMENTS
YARN-285. Added a temporary plugin interface for RM to be able to redirect
to JobHistory server for apps that it no longer tracks. (Derek Dagit via
vinodkv)
OPTIMIZATIONS
BUG FIXES
YARN-188. Coverage fixing for CapacityScheduler (Aleksey Gorshkov via
bobby)
YARN-214. RMContainerImpl does not handle event EXPIRE at state RUNNING
(jeagles via bobby)
YARN-151. Browser thinks RM main page JS is taking too long
(Ravi Prakash via bobby)
YARN-204. test coverage for org.apache.hadoop.tools (Aleksey Gorshkov via
bobby)
YARN-251. Proxy URI generation fails for blank tracking URIs (Tom White
via jlowe)
YARN-258. RM web page UI shows Invalid Date for start and finish times
(Ravi Prakash via jlowe)
YARN-266. RM and JHS Web UIs are blank because AppsBlock is not escaping
string properly (Ravi Prakash via jlowe)
YARN-280. RM does not reject app submission with invalid tokens
(Daryn Sharp via tgraves)
YARN-225. Proxy Link in RM UI thows NPE in Secure mode
(Devaraj K via bobby)
YARN-293. Node Manager leaks LocalizerRunner object for every Container
(Robert Joseph Evans via jlowe)
YARN-50. Implement renewal / cancellation of Delegation Tokens
(Siddharth Seth via tgraves)
YARN-320. RM should always be able to renew its own tokens.
(Daryn Sharp via sseth)
YARN-325. RM CapacityScheduler can deadlock when getQueueInfo() is
called and a container is completing (Arun C Murthy via tgraves)
YARN-334. Maven RAT plugin is not checking all source files (tgraves)
YARN-354. WebAppProxyServer exits immediately after startup (Liang Xie via
jlowe)
Release 0.23.5 - 2012-11-28
INCOMPATIBLE CHANGES
NEW FEATURES
IMPROVEMENTS
YARN-161. Fix multiple compiler warnings for unchecked operations in YARN
common. (Chris Nauroth via vinodkv)
YARN-43. Fix TestResourceTrackerService to not depend on test order and thus
pass on JDK7. (Thomas Graves via vinodkv)
YARN-32. Fix TestApplicationTokens to not depend on test order and thus pass
on JDK7. (vinodkv)
YARN-186. Coverage fixing LinuxContainerExecutor (Aleksey Gorshkov via
bobby)
YARN-216. Remove jquery theming support. (Robert Joseph Evans via jlowe)
OPTIMIZATIONS
BUG FIXES
YARN-163. Retrieving container log via NM webapp can hang with multibyte
characters in log (jlowe via bobby)
YARN-174. Modify NodeManager to pass the user's configuration even when
rebooting. (vinodkv)
YARN-177. CapacityScheduler - adding a queue while the RM is running has
wacky results (acmurthy vai tgraves)
YARN-178. Fix custom ProcessTree instance creation (Radim Kolar via bobby)
YARN-180. Capacity scheduler - containers that get reserved create
container token to early (acmurthy and bobby)
YARN-139. Interrupted Exception within AsyncDispatcher leads to user
confusion. (Vinod Kumar Vavilapalli via jlowe)
YARN-165. RM should point tracking URL to RM web page for app when AM fails
(jlowe via bobby)
YARN-159. RM web ui applications page should be sorted to display last app
first (tgraves via bobby)
YARN-166. capacity scheduler doesn't allow capacity < 1.0 (tgraves via
bobby)
YARN-189. Fixed a deadlock between RM's ApplicationMasterService and the
dispatcher. (Thomas Graves via vinodkv)
YARN-202. Log Aggregation generates a storm of fsync() for namenode
(Kihwal Lee via bobby)
YARN-201. Fix CapacityScheduler to be less conservative for starved
off-switch requests. (jlowe via acmurthy)
YARN-206. TestApplicationCleanup.testContainerCleanup occasionally fails.
(jlowe via jeagles)
YARN-212. NM state machine ignores an APPLICATION_CONTAINER_FINISHED event
when it shouldn't (Nathan Roberts via jlowe)
YARN-219. NM should aggregate logs when application finishes. (bobby)
Release 0.23.4
INCOMPATIBLE CHANGES
NEW FEATURES
IMPROVEMENTS
Change package of YarnClient to org.apache.hadoop. (Bikas Saha via vinodkv)
YARN-108. FSDownload can create cache directories with the wrong
permissions (Jason Lowe via bobby)
YARN-57. Allow process-tree based resource calculation et al. to be
pluggable to support it on multiple platforms. (Radim Kolar via acmurthy)
OPTIMIZATIONS
BUG FIXES
YARN-88. DefaultContainerExecutor can fail to set proper permissions.
(Jason Lowe via sseth)
YARN-106. Nodemanager needs to set permissions of local directories (jlowe
via bobby)
Release 0.23.3
INCOMPATIBLE CHANGES
NEW FEATURES
IMPROVEMENTS
OPTIMAZATIONS
BUG FIXES
YARN-14. Symlinks to peer distributed cache files no longer work
(Jason Lowe via bobby)
YARN-25. remove old aggregated logs (Robert Evans via tgraves)
YARN-27. Failed refreshQueues due to misconfiguration prevents further
refreshing of queues (Arun Murthy via tgraves)
MAPREDUCE-4323. NM leaks filesystems (Jason Lowe via jeagles)
YARN-39. RM-NM secret-keys should be randomly generated and rolled every
so often. (vinodkv and sseth via sseth)
YARN-31. Fix TestDelegationTokenRenewer to not depend on test order so as to
pass tests on jdk7. (Thomas Graves via vinodkv)
YARN-63. RMNodeImpl is missing valid transitions from the UNHEALTHY state
(Jason Lowe via bobby)
YARN-60. Fixed a bug in ResourceManager which causes all NMs to get NPEs and
thus causes all containers to be rejected. (vinodkv)
YARN-66. aggregated logs permissions not set properly (tgraves via bobby)
YARN-68. NodeManager will refuse to shutdown indefinitely due to container
log aggregation (daryn via bobby)
YARN-87. NM ResourceLocalizationService does not set permissions of local
cache directories (Jason Lowe via tgraves)