f7f078b | Arunachalam Thirupathi | 16 August 2014, 14:19:06 UTC | Directly write to socket instead of NIO Directly write to socket instead of NIO doing the write for you. This will avoid one round trip for the selector which will improve the latency and also throughput. pending VPL perf lab runs. Admin still uses NIO for streaming writes as I do not have time to verify it and I see less return of value for Admin writes. | 06 September 2014, 01:13:53 UTC |
b8dbbe7 | ARUNACHALAM THIRUPATHI | 04 September 2014, 07:31:13 UTC | ZoneShrinkage endToendTest fails Zone is shrinked when the nodes receive get/put traffic. The bootstrap URL is set to node 1 and node 0 is removed at the end of zone shrinkage. Zone shrinkage metadata is written to nodes in order. So first node 0 receives the update, that it is no longer part of the cluster. But before node 1 is updated, the bootstrap code retries to bootstrap from node 1. This is a race condition between admin client setting correct metadata on all nodes and the client threads refreshing the old value from other nodes. The problem is update metadataversion properties fail as the node 0 is not longer a part of the cluster. The race condition will be avoided if the bootstrap url and the node in which we first write the metadata change are one and the same. This is what this specific fix does, by making sure node 0 is not removed during zone shrinkage. | 05 September 2014, 14:55:41 UTC |
e06bd37 | Arunachalam Thirupathi | 05 September 2014, 00:08:04 UTC | Releasing Voldemort 1.8.16 Releasing voldemort 1.8.16 | 05 September 2014, 00:08:04 UTC |
a0ca5c8 | Arunachalam Thirupathi | 04 September 2014, 23:36:23 UTC | QuotaException causes delete failure 1) When QuotaException is thrown by delete, it fails the delete instead of Quorum to decide the failure. 2) Added unit test to cover some part of the delete test case. | 05 September 2014, 00:02:45 UTC |
19a97db | Xu Ha | 09 May 2014, 23:23:32 UTC | bdb native backup fixups 1. add message for native backup status 2. add bdb env config param: bdb.recovery.force.checkpoint | 03 September 2014, 19:30:30 UTC |
c2c1473 | Felix GV | 02 September 2014, 23:28:40 UTC | Releasing Voldemort 1.8.15 | 02 September 2014, 23:28:40 UTC |
60e99e4 | Felix GV | 30 August 2014, 00:17:42 UTC | Made the read quota take all GET_ALL keys into account | 30 August 2014, 00:17:42 UTC |
3e16d36 | Arunachalam Thirupathi | 28 August 2014, 22:48:01 UTC | Merge Read and Write Quota Combined GET and GETALL to GET Combined PUT and DELETE to PUT It would be cleaner if we rename the new GET to READ and PUT to WRITE But it might break the backward compatibility so leaving it like this for now. | 29 August 2014, 00:10:49 UTC |
d5c048b | Felix GV | 28 August 2014, 20:53:16 UTC | Merge branch 'master' of https://github.com/voldemort/voldemort | 28 August 2014, 20:53:16 UTC |
c080cfc | Felix GV | 09 August 2014, 00:16:19 UTC | Removed old parent delegating code in StoreStats which would cause double counting. Conflicts: src/java/voldemort/store/stats/StoreStats.java | 28 August 2014, 17:52:24 UTC |
d7d7349 | ARUNACHALAM THIRUPATHI | 20 August 2014, 06:35:14 UTC | Slop for delete operations Problems: 1) Exception handling of delete is very different at 4 places ( on normal response, required failure, quorum failure) and after pipeline is finished. 2) The exceptions are reported again and again ( They are not removed from the map). 3) Some places ignore obsoleteVersionException, some others report it. 4) There is a zombie state abort, there is no way to reach this state. 5) Multiple slops could be sent, because of the issue 2. When the pipeline is aborted, no slops could be sent. 6) Refactored QuotaLimitingStore test to add delete test cases 7) Combined the PUT and GET quotas into 2 quotas. Solution: Defined a common method, so that all 4 places call into the same method. Only the condition for calling is different. Race conditions still exist, after zone failure check but before pipeline finishes the exception will go missing, slops will not happen, but the chances are reduced. Got rid of the state PerformDeletedHintHandoff as QuotaExceededExceptions will not be reported as failures. Now doing it in place. | 27 August 2014, 02:41:21 UTC |
58ab1dc | ARUNACHALAM THIRUPATHI | 27 August 2014, 01:40:09 UTC | Merge pull request #208 from readams/master Fix bitrot and make it build again with newer toolchain | 27 August 2014, 01:40:09 UTC |
453ddc1 | Rob Adams | 27 August 2014, 01:05:22 UTC | Fix the lib.private section of pkg-config file | 27 August 2014, 01:05:22 UTC |
37e18e7 | Rob Adams | 18 August 2014, 20:51:50 UTC | Fix bitrot and make it build again with newer toolchain | 27 August 2014, 00:58:31 UTC |
8724f64 | Siddharth Singh | 27 August 2014, 00:37:25 UTC | Releasing Voldemort 1.8.14 | 27 August 2014, 00:37:25 UTC |
ed86e11 | Felix GV | 26 August 2014, 23:20:21 UTC | Upgraded to Tehuti 0.5 Tehuti 0.5 includes a fix about Histograms not clearing their samples properly. Also added some trace logging code in StoreStats. | 26 August 2014, 23:20:21 UTC |
b762a80 | ARUNACHALAM THIRUPATHI | 25 August 2014, 23:12:41 UTC | Merge pull request #212 from arunthirupathi/CleanUpRebalance DataCleanup job should not run during rebalance | 25 August 2014, 23:12:41 UTC |
3f0fc7b | Xu Ha | 25 August 2014, 22:40:42 UTC | Releasing Voldemort 1.8.13 | 25 August 2014, 22:40:42 UTC |
eaf9b6f | ARUNACHALAM THIRUPATHI | 25 August 2014, 22:18:12 UTC | Merge pull request #210 from arunthirupathi/deleteGetVersion Delete should use GetVersion instead of Get | 25 August 2014, 22:18:12 UTC |
0371771 | Arunachalam Thirupathi | 24 August 2014, 16:41:13 UTC | Log warning for GET ALL and DELETE Log warnings for other than GET and PUT. | 24 August 2014, 16:41:13 UTC |
90cd6a8 | ARUNACHALAM THIRUPATHI | 24 August 2014, 14:33:31 UTC | Merge pull request #214 from zhongjiewu/example-config-cleanup Migrate from stores.xml to STORES folder in all example configs | 24 August 2014, 14:33:31 UTC |
62aa593 | Zhongjie Wu | 24 August 2014, 00:10:57 UTC | Migrate from stores.xml to STORES folder in all example configs | 24 August 2014, 00:14:38 UTC |
9bbe12a | ARUNACHALAM THIRUPATHI | 23 August 2014, 06:22:20 UTC | Merge pull request #213 from zhongjiewu/better-example Polish up the Java Client example | 23 August 2014, 06:22:20 UTC |
5307345 | Zhongjie Wu | 23 August 2014, 00:58:14 UTC | Polish up the Java Client example | 23 August 2014, 00:59:16 UTC |
8281397 | Arunachalam Thirupathi | 22 August 2014, 23:01:38 UTC | DataCleanup job should not run during rebalance 1) There is a check to not start Data Cleanup job during rebalance, but after it is started, there is no periodic check. This causes the data cleanup to compete with rebalance. 2) Check every 10,000 entries scanned to make sure that there is no rebalance in progress. 3) Added error condition around the data retention to make sure that if it is set to 0 or lesser days the value is ignored and an error is logged. Previously if it is set to 0, it will wipe out your entire data. | 22 August 2014, 23:01:38 UTC |
d8c9353 | ARUNACHALAM THIRUPATHI | 22 August 2014, 21:20:06 UTC | Merge pull request #211 from arunthirupathi/node0Issue Tools depend on Node 0 to be available | 22 August 2014, 21:20:06 UTC |
b1b68a3 | Arunachalam Thirupathi | 22 August 2014, 19:35:20 UTC | Delete should use GetVersion instead of Get 1) Currently Delete uses Get instead of GetVersion before doing delete. This can cause the Get to count towards Quota and is also unnecessary overhead when you want to delete things. 2) Now GetVersion is attempted and if multiple versions are encountered, then a Get is done to do the read repair. 3) Put uses duplicated code, the code is consolidated. Made the methods private. I think I should replace the GetVersion method with the new method I wrote, but not sure what is the use case and saving that for a later date. | 22 August 2014, 19:35:20 UTC |
105915d | Arunachalam Thirupathi | 22 August 2014, 00:06:13 UTC | Tools depend on Node 0 to be available Added a new override for getRemoteStoreDefList which takes no parameters and identifies one of the nodes, for it to be used. Fixed all the tools (not used in production code path but could be used by SREs) code, which used NodeId 0 as default to use the new overload. VoldemortMultiStoreBuildAndPushJob seems to have the issue, but not sure, whether it is used in production, so just added the comment. | 22 August 2014, 00:06:13 UTC |
c07b777 | Xu Ha | 14 August 2014, 23:40:59 UTC | fix duplicate error message bug and add large value size test for coordinator | 18 August 2014, 21:23:59 UTC |
6644d32 | ARUNACHALAM THIRUPATHI | 14 August 2014, 21:30:23 UTC | Merge pull request #207 from FelixGV/master Fixed the Show Spurious Values test... | 14 August 2014, 21:30:23 UTC |
7e774cb | Felix GV | 14 August 2014, 17:18:38 UTC | Fixed the Show Spurious Values test... | 14 August 2014, 20:26:32 UTC |
265b4c9 | Siddharth Singh | 14 August 2014, 16:58:18 UTC | Rename variable | 14 August 2014, 17:09:38 UTC |
25b1cf7 | Arunachalam Thirupathi | 14 August 2014, 00:41:34 UTC | Remove tehuit from lib as it is auto generated The jar should be copied from private-lib, for some weird reason I thought the lib is checked in and hence I checked in but it is not required. | 14 August 2014, 00:41:34 UTC |
28e19b2 | ARUNACHALAM THIRUPATHI | 14 August 2014, 00:40:52 UTC | Merge pull request #206 from arunthirupathi/zoneCheck Zone check | 14 August 2014, 00:40:52 UTC |
6d8c05d | Arunachalam Thirupathi | 08 August 2014, 06:22:24 UTC | Zone check, store name in error, unit tests 1) When Server fails to start because of an invalid store, the store is logged in the error message. So that it is actionable. 2) Added checks to zone proximity list to avoid same zone and duplicate zone ids. 3) Modified Zones from LinkedList to ArrayList, as get operation is more efficient in ArrayList than the LinkedList. 4) Refactor the common code in zone calculations to common functions. 5) Added unit tests to cover the new checks added to zone proximity list. | 13 August 2014, 22:47:55 UTC |
044e96f | FelixGV | 12 August 2014, 01:56:46 UTC | Merge pull request #205 from FelixGV/master Releasing Voldemort 1.8.12 | 12 August 2014, 01:56:46 UTC |
f29a973 | Felix GV | 12 August 2014, 01:30:04 UTC | Releasing Voldemort 1.8.12 | 12 August 2014, 01:30:04 UTC |
56e199d | Arunachalam Thirupathi | 12 August 2014, 00:49:45 UTC | Tools are failing because of the missing files Added tehuti to the lib directory. | 12 August 2014, 00:49:45 UTC |
f5e886d | Arunachalam Thirupathi | 12 August 2014, 00:39:00 UTC | Merge conflict was incorrectly resolved Fixed the merge conflict | 12 August 2014, 00:39:00 UTC |
5b7f23a | Arunachalam Thirupathi | 12 August 2014, 00:32:38 UTC | Merge conflict on StatsTest Resolved the merge conflict | 12 August 2014, 00:32:38 UTC |
2df417d | Felix GV | 07 August 2014, 18:10:17 UTC | Upgraded to latest Tehuti, 0.4 | 12 August 2014, 00:12:21 UTC |
d8fd80a | Felix GV | 06 August 2014, 00:49:12 UTC | Various fixes for the stats code... | 12 August 2014, 00:12:21 UTC |
9f5b30e | Felix GV | 05 August 2014, 22:10:22 UTC | Fixed bugs related to ClientSocketStatsTest... | 12 August 2014, 00:12:20 UTC |
651c67a | Felix GV | 05 August 2014, 20:04:57 UTC | Fixed a problem with RequestCounter.getNumEmptyResponses(). It required a SampledTotal, not a SampledCount. | 12 August 2014, 00:12:20 UTC |
56daa0c | Felix GV | 05 August 2014, 18:53:42 UTC | Fixed a problem in StoreStatsJmxTest where Tehuti's Max would return NEGATIVE_INFINITY. | 12 August 2014, 00:12:20 UTC |
add55a9 | Felix GV | 05 August 2014, 01:49:29 UTC | Fixed a test in StatsTest to use the new RequestCounter API. | 12 August 2014, 00:12:20 UTC |
53d6b0e | Felix GV | 05 August 2014, 01:37:57 UTC | Initial commit of the Tehuti stats integration. Some unit tests are failing. Probably a few bugs need squashing... | 12 August 2014, 00:12:20 UTC |
c9a5397 | Siddharth Singh | 31 July 2014, 02:05:16 UTC | Skewed reporting | 12 August 2014, 00:12:19 UTC |
c3773f4 | Zhongjie Wu | 11 August 2014, 22:01:01 UTC | Fix a vadmin error | 11 August 2014, 22:01:06 UTC |
6cab7b1 | Travis Gockel | 02 August 2014, 16:46:05 UTC | C++ Client: bump maintenance version number. | 07 August 2014, 04:33:55 UTC |
e3c4375 | Travis Gockel | 31 July 2014, 18:36:20 UTC | Makes Store::get and Store::getName const functions. Also changes InconsistencyResolvingStore, RoutedStore and SocketStore implementations to be const. | 07 August 2014, 04:33:55 UTC |
559d202 | Travis Gockel | 31 July 2014, 17:48:35 UTC | C++ client: Changes VersionedValue::getVersion to return a const Version*. By returning a non-const Version*, the semantic const-correctness was broken. You could not modify the pointer, but that didn't matter. | 07 August 2014, 04:33:54 UTC |
45bc4ea | Travis Gockel | 31 July 2014, 17:01:00 UTC | C++ client: Adds VOLDEMORT_ prefix to all include guards. This should minimize conflicts with other libraries. | 07 August 2014, 04:33:54 UTC |
9f40015 | Travis Gockel | 31 July 2014, 16:36:17 UTC | Changes exception message in Cluster::getNodeById to use boost::format. The previous code would cause a segmentation fault if nodeId was greater than the string length, and the message now actually includes the ID. | 07 August 2014, 04:33:54 UTC |
a51f328 | Travis Gockel | 31 July 2014, 16:31:58 UTC | Changes Cluster constructor to take string by const reference. | 07 August 2014, 04:33:54 UTC |
f3c6fdb | Arunachalam Thirupathi | 04 August 2014, 19:42:08 UTC | Four node and two node configs two configurations for local testing. I found it helpful for testing some of the scenarios. Editing them on multiple machines is painful and others could use this for testing. | 04 August 2014, 21:15:48 UTC |
acd1117 | Arunachalam Thirupathi | 04 August 2014, 19:40:29 UTC | gradle build fails on mac some of the mac is forced to use Java 8 and Scala plugin does on Gradle 1.12 does not work with Java 8. Gradle has fixed the issue on 2.0 version. Added a task to cleanup the config directory files so that changes to stores.xml will take effect. Scala build intermittently fails with the error on mac FAILURE: Build failed with an exception. What went wrong: Could not resolve all dependencies for configuration ':zinc'. > Could not find org.scala-lang:scala-library:2.10.2. This is caused by incremental zinc compilation. Given the fact that scala is very rarely used the intermittent failure is very hard to workaround. Added a task for cleaning up config folder so that changes in stores.xml will take effect. | 04 August 2014, 21:15:48 UTC |
ec2fcfd | Siddharth Singh | 04 August 2014, 18:30:01 UTC | Reuse objectmapper object to make vector clock serialization faster | 04 August 2014, 18:32:49 UTC |
0a0ef0c | Arunachalam Thirupathi | 01 August 2014, 22:02:49 UTC | Releasing voldemort 1.8.11 | 01 August 2014, 22:02:49 UTC |
970877c | Arunachalam Thirupathi | 30 July 2014, 17:52:00 UTC | Exceptions goes uncounted Get does parallel requests first and if it fails/times move on to the serial requests. Exceptions are tracked if the parallel requests fail or timesout after the pipeline is finished. But if the Serial requests are processing then the exception goes unnoticed. This causes the node to stay available for a long time affecting the latency and error rates. This fix reduces the time window an exception goes unnoticed. I thought about moving the error handling inside the callback and not have it at 2 places. But not sure what will be the perf impact of the error tracking code, could be negligible but saving that for a later time. | 30 July 2014, 17:52:00 UTC |
51b2395 | Arunachalam Thirupathi | 30 July 2014, 17:41:25 UTC | ClientRequest Executor fix and log fixes 1) Protocol negotiation request does not have a timeout associated with it . So if the connection fails the Input/Output buffers are leaked and cause Out Of Memory errors. Setting the timeout for the Protocol negotiation requests as the caller expects a timeout. No new logs are introduced. Only the existing log is made more informative. 2) When protocol negotiation timesout, an obscure error message is thrown, IllegalStateException. This happens as the code tries to read the result in case of timeout. Handle this case and report a meaningful error. 3) Serial request times out on a connection, error message does not mention the socket that timed out. 4) Fix other minor logging information. | 30 July 2014, 17:41:25 UTC |
5708c0c | Arunachalam Thirupathi | 29 July 2014, 20:59:53 UTC | Threshold failure detector issue 1) Threshold failure detector, marks a node as available when it receives a non-catastrophic error after configured number of catastrophic errors. When a node goes unavailable, generally first there will be lots of connectExceptions ( catastrophic) and timeouts for already established connections. This causes failure detector to treat a node that is being down as up and affects the client latency as all the get waits for the connection timeout to happen and goes for the serial requests. 2) Threshold failure detector, marks a node as unavailable after a window rolls over. Threshold failure detector tracks successes/failures in window ( default 5 minutes) and if the successes drop below a configured ( defaul 95 ) percentage and if there are more errors than configured it marks the node as down. When the window rolls over, and if the first request succeeds or when a node comes back up the failure count is not reset. This causes an available node to be marked as down. 3) Enhance the unit tests to cover the above 2 cases. 4) Dump the statistics when a node goes down to reason about it in the logs. The new code is written with the following reasoning. Do the book keeping first. Make decision next. The new code intentionally does not reset the catastrophic errors on a window roll over as it will be reset by the first success anyway. The node can still flap when the failure is above minimum and the percentage oscillates around the configured percentage. But I don't see any good workaround and nor it was an actual issue on 20+ repetitions I created for an internal repro. So leaving them as it is. Previous code mixed both of these and lead to many issues. The previous code left many loose ends ( Boolean represented in int, Generic set methods, when it required reset, Set methods at individual variable level, when only reset method is required, copy pasted code). PS: There is a more serious third issue, Selector drains the parallel queue. When the request requires connection establishment Selector handles the connection too. If the node is dead Selector is going to wait configured time for connection to timeout ( default 5 seconds) . In this time Selector is not pumping read/write and hence all these requests eventually time out. We are discussing the potential issues for this fix and will address this in the next fixes. There are many other minor issues I uncovered as well. | 29 July 2014, 21:17:38 UTC |
4fd732f | Siddharth Singh | 25 July 2014, 23:52:07 UTC | Expose http decoder parameters as config | 28 July 2014, 19:53:41 UTC |
a86898e | Bhavani Sudha Saktheeswaran | 23 July 2014, 23:49:52 UTC | Releasing Voldemort 1.8.10 | 23 July 2014, 23:49:52 UTC |
ee09314 | Bhavani Sudha Saktheeswaran | 16 July 2014, 05:40:35 UTC | Register slops if async writes fail due to quota Fix Async puts to succeed even if quota exceeded Please enter the commit message for your changes. Lines starting Adding unit test for Quota failed Parallel puts Fixing unit tests for Async write quota failures Updating PipelineRoutedStats JMXGetters for QuotaExceededException And minor code chagnes | 23 July 2014, 23:34:51 UTC |
493f500 | Siddharth Singh | 16 July 2014, 20:07:29 UTC | Remove dated inaccurate comment | 16 July 2014, 20:08:03 UTC |
30185d4 | Bhavani Sudha Saktheeswaran | 02 July 2014, 22:05:25 UTC | Releasing Voldemort 1.8.9 | 02 July 2014, 22:05:25 UTC |
0a9b7ed | Bhavani Sudha Saktheeswaran | 02 July 2014, 20:49:38 UTC | minor code review comments | 02 July 2014, 20:49:38 UTC |
28f5d9f | Bhavani Sudha Saktheeswaran | 02 July 2014, 18:40:52 UTC | Adding schema backwards comaptibility in AdminClient for these methods: * updateRemoteMetadata * updateRemoteMetadataPair * addStore | 02 July 2014, 18:40:52 UTC |
a42ac35 | Bhavani Sudha Saktheeswaran | 02 July 2014, 00:18:37 UTC | Adding backwards compatibility checks in AdminClient For the methods: * fetchAndUpdateRemoteStore(int nodeId, List<StoreDefinition> updatedStores) * updateRemoteStoreDefList(int nodeId, List<StoreDefinition> storesList) so the exceptions are thrown even before hitting the server. | 02 July 2014, 00:18:37 UTC |
7ec06ad | Zhongjie Wu | 25 June 2014, 00:20:17 UTC | Add schema backward compatibility check when updating store definitions | 25 June 2014, 00:20:17 UTC |
cd4c627 | Bhavani Sudha Saktheeswaran | 24 June 2014, 20:50:41 UTC | Releasing Voldemort 1.8.8 | 24 June 2014, 20:50:41 UTC |
8dbdb63 | Bhavani Sudha Saktheeswaran | 23 June 2014, 21:53:23 UTC | Incorporating review comments Changing timeout for HttpClientFactory shutdown to 10 seconds | 24 June 2014, 20:09:41 UTC |
ca6d63c | Bhavani Sudha Saktheeswaran | 23 June 2014, 21:15:03 UTC | move transport client instantiation to getRawStore adding timeout to HttpClientFactory shutdown added minor log messages | 24 June 2014, 20:09:41 UTC |
de7cae0 | singhsiddharth | 24 June 2014, 19:17:13 UTC | Merge pull request #199 from arunthirupathi/eclipseJunit Junit tests are not running inside eclipse | 24 June 2014, 19:17:13 UTC |
7c9af85 | Arunachalam Thirupathi | 24 June 2014, 17:12:08 UTC | Junit tests are not running inside eclipse Java builder property is not set on the .project file which causes all the tests to fail with the ClassNotFoundException Some test uses GetContent which causes CoordinatorRestAPITest to fail. | 24 June 2014, 17:12:08 UTC |
11c8664 | vinoth chandar | 20 June 2014, 20:08:56 UTC | Update CONTRIBUTORS Add Arun and Xu | 20 June 2014, 20:08:56 UTC |
30b0c32 | Siddharth Singh | 19 June 2014, 21:45:18 UTC | Wrap mime multipart into a mime message to ensure headers are updated | 19 June 2014, 22:46:18 UTC |
674b152 | Xu Ha | 19 June 2014, 22:22:44 UTC | add socket connection refresh message in AdminClient | 19 June 2014, 22:22:44 UTC |
2d531e2 | Chinmay Soman | 19 June 2014, 18:44:14 UTC | Releasing Voldemort 1.8.7 with the Quota metrics fix | 19 June 2014, 18:44:14 UTC |
9065ec1 | ARUNACHALAM THIRUPATHI | 19 June 2014, 18:36:49 UTC | Merge pull request #198 from icefury71/mbean_quota_fix Registering individual store quota stats instead of aggregate to the Mbe... | 19 June 2014, 18:36:49 UTC |
3d63974 | Arunachalam Thirupathi | 19 June 2014, 17:41:51 UTC | hudson run fails to copy mysql 1) There is not libs directory anymore so hudson run fails to copy the mysql.jar to the lib directory. Added the mysql as testRuntime so the jar will be downloaded automatically 2) Added war task to update the documentation | 19 June 2014, 17:43:40 UTC |
cf92ed7 | Chinmay Soman | 19 June 2014, 00:21:45 UTC | Registering individual store quota stats instead of aggregate to the Mbean server | 19 June 2014, 17:25:32 UTC |
6dc8505 | ARUNACHALAM THIRUPATHI | 19 June 2014, 06:25:55 UTC | add eclipse and idea projects to gitignore eclipse and idea projects will have the lib paths different for each person, as they will point to the jars from the gradle cache. | 19 June 2014, 06:25:55 UTC |
f0ee03d | ARUNACHALAM THIRUPATHI | 18 June 2014, 19:32:38 UTC | remove .classpath and .project gradle eclipse updates teh existing classpath and project and leaves teh classpath is a very inconsistent state. So removing these files and the users can generate them when required. Will fix the .gitignore to add these files so that they dont show up. | 18 June 2014, 19:32:38 UTC |
71e29a7 | Arunachalam Thirupathi | 18 June 2014, 15:18:30 UTC | eclipse import is broken 1) src/java is defined multiple times which causes the problem 2) outputDir is set to bin, which erases all the shell scripts | 18 June 2014, 15:18:30 UTC |
861dbe4 | Arunachalam Thirupathi | 04 June 2014, 00:31:17 UTC | use maven central for gradle dependencies 1) The jars are directly consumed from libs directory which makes upgrading the dependencies harder. Also in some cases the transitive dependencies are not copied correctly which causes issues. 2) Seperated the libs to private-lib and public-lib. private-lib will contain jars for which I could not find the same version in the maven central repository. The public-lib are available in maven central and are not dependent upon by gradle. But still kept for ant compatibility. 3) As part of the build (ant/gradle) all the dependent libs are copied into the libs. This is to make the bin/*.sh work. Also there is some EasyVPL code which depends on the jar in library. 4) Since the classpath is going to be different both the eclipse and idea will break. But after the download you can run ./gradlew eclipse and ./gradlew idea to regenerate the correct project/ classpath to make it work again. | 13 June 2014, 23:49:28 UTC |
99e7207 | ARUNACHALAM THIRUPATHI | 11 June 2014, 22:01:34 UTC | Merge pull request #188 from d4mian/minor-changes Comment typos, duplicate code removal and avoid unnecessary fetch of column key_ from the database. | 11 June 2014, 22:01:34 UTC |
e639011 | Chinmay Soman | 11 June 2014, 21:31:32 UTC | Releasing Voldemort 1.8.6 containing Quota (per store instead of aggregate) fix. | 11 June 2014, 21:31:32 UTC |
3b4dd8d | ARUNACHALAM THIRUPATHI | 11 June 2014, 17:57:10 UTC | Merge pull request #196 from icefury71/quota_per_store_stats_fix Fix for QuotaLimitingStore: Passing storestats for the current store ins... | 11 June 2014, 17:57:10 UTC |
25800b8 | singhsiddharth | 11 June 2014, 17:09:37 UTC | Merge pull request #194 from arunthirupathi/PoolSizeRename RestClient renamed poolsize key | 11 June 2014, 17:09:37 UTC |
6b4c17f | Chinmay Soman | 11 June 2014, 16:57:08 UTC | Fix for QuotaLimitingStore: Passing storestats for the current store instead of the aggregate | 11 June 2014, 16:57:08 UTC |
1b31ab8 | Arunachalam Thirupathi | 10 June 2014, 22:26:39 UTC | log4j.properties is included in the voldemort jar While refactoring the consultant moved some code and it caused binary level difference on the vold jar. I reverted the change for now. Since it is reverting to the old state, skipping code review for now. | 10 June 2014, 22:35:41 UTC |
a2adb70 | Arunachalam Thirupathi | 10 June 2014, 17:03:37 UTC | RestClient renamed poolsize key RestClient does not have POOL_SIZE_KEY anymore. From the source it is probably renamed to HTTP_POOL_SIZE. The previous constant value was poolsize and the new constant is http.poolsize. Upgrading the R2, Data, Pegasus libraries along with this change. | 10 June 2014, 17:03:37 UTC |
f41a40f | Xu Ha | 03 June 2014, 18:21:13 UTC | add optional identifier string for StoreClientFactory 1. fix invalid package name in voldemort-coordinator.sh 2. replace jmxId with identifierString | 10 June 2014, 00:28:41 UTC |
89389c3 | Arunachalam Thirupathi | 06 June 2014, 18:10:07 UTC | Release voldemort 1.8.5 Gradle related changes - empty settings file - empty voldemort-contrib directory - gradle wrapper to fetch required version | 06 June 2014, 18:10:20 UTC |
f0f2ac8 | ARUNACHALAM THIRUPATHI | 06 June 2014, 18:01:32 UTC | Merge pull request #193 from arunthirupathi/voldemort-contrib Dummy directory for injecting custom gradle behavior | 06 June 2014, 18:01:32 UTC |
d540995 | Arunachalam Thirupathi | 06 June 2014, 17:54:38 UTC | Dummy directory for injecting custom gradle behavior This is a place holder directory for injecting custom gradle behavior. This directory is not directly used in the open source gradle builds. The open source project produces 2 artifacts voldemort.jar and voldemort-contrib.jar. This is not gradle friendly as in gradle if you are producing more than one jar from the same module exporting ivy dependencies and other becomes very trickier. This allows an upstream wrapper to inject custom behavior. For example upstream wrapper can define a project(':voldemort:voldemort-contrib') and define all the gradle behavior dependencies, black box build . Theoretically we could have used the contrib directory but that will result in lot more confusion as that directory produces some artifacts and upstream could want a different behavior. | 06 June 2014, 17:54:38 UTC |
7ee12b1 | ARUNACHALAM THIRUPATHI | 05 June 2014, 16:02:07 UTC | Merge pull request #192 from szczepiq/master Added empty settings.gradle so that Gradle does not search it in parent dirs. | 05 June 2014, 16:02:07 UTC |
9d0a32d | Szczepan Faber | 05 June 2014, 14:22:43 UTC | Added empty settings.gradle so that Gradle does not search for it in parent dirs. | 05 June 2014, 14:22:43 UTC |