7df580e | Dan LaRocque | 26 December 2013, 00:00:55 UTC | [maven-release-plugin] prepare release 0.4.2 | 26 December 2013, 00:00:55 UTC |
4b4fecb | Dan LaRocque | 25 December 2013, 23:59:04 UTC | Update wikidoc commit pointer | 25 December 2013, 23:59:04 UTC |
420ffeb | Dan LaRocque | 25 December 2013, 23:58:34 UTC | Make maven-gpg-plugin use gpg-agent | 25 December 2013, 23:58:34 UTC |
2d223e0 | Dan LaRocque | 25 December 2013, 17:47:33 UTC | Overwrite UPGRADE.textile with wiki copy | 25 December 2013, 17:47:33 UTC |
b45c5ce | Dan LaRocque | 24 December 2013, 02:00:57 UTC | Add 0.4.1 to titan.compatible.versions | 24 December 2013, 02:00:57 UTC |
eb9ec91 | Dan LaRocque | 24 December 2013, 01:55:11 UTC | Copying backlogged changes from wiki to CHANGELOG Some updates to the Release-Notes page of the wiki were never propagated back into CHANGELOG.textile (including the entire set of 0.3.2 release notes). They should be in sync again now. | 24 December 2013, 01:56:47 UTC |
1ff3d2d | Dan LaRocque | 24 December 2013, 01:49:56 UTC | Updated CHANGELOG for 0.4.2 | 24 December 2013, 01:49:56 UTC |
ef5a136 | Dan LaRocque | 23 December 2013, 22:23:56 UTC | Add Metrics counter of getVertex(long) calls Part of #402 | 23 December 2013, 22:23:56 UTC |
034750d | Dan LaRocque | 23 December 2013, 21:21:24 UTC | Replace AtomicLongs in CachedKCVS with Metrics This is the second half of #485 | 23 December 2013, 21:22:00 UTC |
8acf320 | Dan LaRocque | 23 December 2013, 21:21:06 UTC | Test import cleanup | 23 December 2013, 21:21:06 UTC |
b7e918a | Dan LaRocque | 23 December 2013, 19:50:26 UTC | Change System.err.println to log.debug | 23 December 2013, 19:50:26 UTC |
b8258e1 | Dan LaRocque | 23 December 2013, 19:33:55 UTC | Refactor metrics configuration handling Previous to this commit, GraphDatabaseConfiguration used metrics.enable-basic-metrics while Backend used storage.enable-basic-metrics. There's no good reason for these options to be separate. It's just confusing. The config option wording is also clumsy. In retrospect, it should really have been metrics.enabled. Titan's Metrics support and metrics.* config namespace grew out of the storage backend, but now Metrics covers other stuff in Titan besides storage. To support existing configs, I made GraphDatabaseConfiguration and Backend both consider metrics globally enabled when any one (or more) of the following config options are set to true: * storage.enable-basic-metrics * metrics.enable-basic-metrics * metrics.enabled Storage backend metrics are merged when of the following are true (they are each true by default): * metrics.merge-basic-metrics * storage.merge-basic-metrics | 23 December 2013, 19:33:55 UTC |
1bf0813 | Dan LaRocque | 23 December 2013, 17:18:42 UTC | Update ConsistentKeyLockerTest for Metrics changes ConsistentKeyLockerTest is, by design, extremely sensitive to internals changes in ConsistentKeyLocker or AbstractLocker. The test started failing when I added calls to tx.getConfiguration().getMetricsPrefix() in AbstractLocker. In this commit, I changed ConsistentKeyLocker to allow any number of calls to tx.getConfiguration(). The configuration doesn't hold mutable lock state, and this is the second time I've stumbled over spurious tx.getConfiguration() failures in ConsistentKeyLockerTest, so I think relaxing this constraint is appropriate. | 23 December 2013, 17:18:42 UTC |
9840c19 | Dan LaRocque | 23 December 2013, 16:36:00 UTC | Tweak getSlice and getKeys metrics names * prevent getSlice from attempting to log metrics with an empty-string name prefix when metrics are disabled * rename com.thinkaurelius.titan.getKeys.iterator.* to com.thinkaurelius.titan.stores.getKeys.iterator.* to conform with the naming pattern of every other storage metric | 23 December 2013, 16:36:00 UTC |
a179f32 | Dan LaRocque | 23 December 2013, 16:33:12 UTC | Rename ExpirationStoreCache metrics When I originally added Metrics counters to this class, I mistakenly set the counter names to just "ExpirationStoreCache". Now the names start with GraphDatabaseConfiguration.METRICS_SYSTEM_PREFIX_DEFAULT + "." + "ExpirationStoreCache". In other words, the metrics in here now start with "com.thinkaurelius.titan.sys.ExpirationStoreCache". | 23 December 2013, 16:33:12 UTC |
b32822c | Dan LaRocque | 23 December 2013, 16:31:37 UTC | Add Metrics counters on locking operations Metrics counters added for calls to and exceptions in the write, check, and delete lock methods of AbstractLocker. ConsistentKeyLocker extends AbstractLocker. A transaction-specific metrics name prefix is honored when set. | 23 December 2013, 16:31:37 UTC |
9396857 | Dan LaRocque | 23 December 2013, 16:30:13 UTC | Reformat broken debug log message Fixes #544 | 23 December 2013, 16:30:13 UTC |
1e0a628 | Dan LaRocque | 23 December 2013, 14:28:33 UTC | Wrap QueryExecutor implementations with Metrics This times and counts (invocations and exceptions counted separately) calls to the methods on implementations of the QueryExecutor interface, when the calling transaction has metrics enabled. Any transaction-specific metrics prefix, if configured, is honored. Time spent in iterators returned by methods on the QueryExecutor interface are *not* covered by these metrics. This is part of #402. | 23 December 2013, 14:28:33 UTC |
2cd3aed | Dan LaRocque | 21 December 2013, 04:43:45 UTC | Replacing AtomicLongs with Metrics Counters Because Metric counters do not support the notion of set/reset/clear, only increment and decrement, ExpirationStoreCache's metrics reset methods are now at best approximate. The reset methods may set their respective counters back to zero, or they set them down to a value slightly above zero. | 21 December 2013, 04:43:45 UTC |
d94d616 | Dan LaRocque | 20 December 2013, 21:09:14 UTC | titan.sh: kill Cassandra only after Rexster dies `titan.sh stop` now waits up to 60 seconds by default (though this is configurable through an environment variable) for Rexster to disappear from the jps process table before attempting to kill Cassandra. This addresses issue 2 of 2 in #538. | 20 December 2013, 21:09:14 UTC |
c67ed62 | Dan LaRocque | 20 December 2013, 20:58:32 UTC | Check for jps in $JAVA_HOME/bin when not on $PATH Addresses problem 1 of 2 in issue #538 | 20 December 2013, 20:58:32 UTC |
5752140 | Matthias Broecheler | 20 December 2013, 09:51:25 UTC | Fixing test case: Oversaw reference to tx when refactoring to use graph throughout. | 20 December 2013, 09:51:25 UTC |
ec9abdf | Matthias Broecheler | 20 December 2013, 08:50:02 UTC | Merge branch 'master' of https://github.com/thinkaurelius/titan | 20 December 2013, 08:50:02 UTC |
a46ce94 | Matthias Broecheler | 20 December 2013, 08:49:26 UTC | Fixes #500. Fixes #535. | 20 December 2013, 08:49:26 UTC |
507a3fb | Matthias Broecheler | 20 December 2013, 07:26:45 UTC | Removed wrong test case. | 20 December 2013, 07:26:45 UTC |
99d550c | Matthias Broecheler | 20 December 2013, 07:25:04 UTC | Fixes #505. Fixes #498 | 20 December 2013, 07:25:04 UTC |
b9cf88e | Matthias Broecheler | 20 December 2013, 03:38:19 UTC | Fixes #499. | 20 December 2013, 03:38:19 UTC |
8918017 | Dan LaRocque | 19 December 2013, 16:24:45 UTC | Expunge uses of `uname -o` from shell scripts The -o option is not POSIX standard and does not exist on OS X (and possibly other platforms). Replaced with calls to just `uname`, without options, which under POSIX must behave the same as `uname -s`. This change is untested. It's intended to address #534, but needs to be verified manually before closing the issue. | 19 December 2013, 16:30:52 UTC |
5f8dbdb | Dan LaRocque | 19 December 2013, 03:27:01 UTC | Avoid unnecessary serialVersionUID warning Replaced an anonymous HashMap-extending class with a call to Guava ImmutableMap.of(). The enclosing class is now warning-free. | 19 December 2013, 03:27:49 UTC |
5005da5 | Dan LaRocque | 19 December 2013, 03:25:07 UTC | Add Astyanax RetryBackoffStrategy configuration Fixes #524 | 19 December 2013, 03:25:35 UTC |
ffa27ea | Matthias Broecheler | 19 December 2013, 03:12:24 UTC | Fixes #507 introducing configuration option: ignore-unknown-index-key | 19 December 2013, 03:12:24 UTC |
a1584af | Matthias Broecheler | 19 December 2013, 01:58:06 UTC | toString on types returns the type's name instead of the vertex identifier. Fixes #525. | 19 December 2013, 01:58:06 UTC |
8f279ee | Matthias Broecheler | 19 December 2013, 01:29:51 UTC | Removing left over clear() method. | 19 December 2013, 01:29:51 UTC |
fe35a70 | Matthias Broecheler | 19 December 2013, 01:09:31 UTC | Fixes #307 | 19 December 2013, 01:09:31 UTC |
f2713c5 | Matthias Broecheler | 19 December 2013, 00:35:49 UTC | Reordered name property creation to be first when creating types so that we don't have a lingering type vertex in case the ensuing exception is thrown and the transaction still committed. | 19 December 2013, 00:35:49 UTC |
afb73d6 | Dan LaRocque | 17 December 2013, 11:44:38 UTC | Merge branch 'master' of github.com:tsecheran/titan into tsecheran-master | 17 December 2013, 11:44:38 UTC |
e50826b | Dan LaRocque | 17 December 2013, 10:39:00 UTC | Merge branch 'fix-for-performance' of github.com:mpouttuclarke/titan into mpouttuclarke-fix-for-performance | 17 December 2013, 10:39:00 UTC |
be0b7c1 | Dan LaRocque | 17 December 2013, 09:56:52 UTC | Merge branch 'issue_508_test' of github.com:MrKeyholder/titan into MrKeyholder-issue_508_test | 17 December 2013, 09:56:52 UTC |
b086337 | Dan LaRocque | 16 December 2013, 19:43:15 UTC | Merge branch 'master' into issue_508 | 16 December 2013, 19:43:15 UTC |
6c2171a | MrKeyholder | 12 December 2013, 04:36:41 UTC | Test case for #508 Resource leak when using 0.4.1 + cassandrathrift and iterating over all vertices | 12 December 2013, 04:36:41 UTC |
d687805 | Matt Pouttu-Clarke | 10 December 2013, 00:07:14 UTC | If there is only one thread, then the threadpool is null and this throws error | 10 December 2013, 00:07:14 UTC |
12e1a1e | Dan LaRocque | 05 December 2013, 05:27:52 UTC | Increase HBase and Cassandra timeouts in testing Changed CONNECTION_TIMEOUT_KEY from 10s to 60s for Cassandra and HBase tests. This was motivated by performance testing on EC2, where truncating column families on Cassandra could take over ten seconds after many iterations of the same test method (truncating CFs after each run). I'm not sure why this operation was slow, but since it's just CF truncation in clearStorage() and so far does not appear to affect querying, I'm just going to bump up the timeout. The 60s value made the timeouts disappear on an EC2 c3.xlarge. | 05 December 2013, 05:27:52 UTC |
88918a9 | Dan LaRocque | 05 December 2013, 05:04:55 UTC | Updated jub-effort.txt for c3.xlarge | 05 December 2013, 05:04:55 UTC |
54c758f | Pavel Yaskevich | 04 December 2013, 21:41:59 UTC | don't hold on to the borrowed connections in Cassandra Thrift iterators | 04 December 2013, 21:41:59 UTC |
b28e83b | tsecheran | 04 December 2013, 19:50:01 UTC | Fix for gremlin.sh will take more than 9 params | 04 December 2013, 19:50:01 UTC |
7bb3f9d | Dan LaRocque | 03 December 2013, 23:36:33 UTC | Allow env var override of perf test target time | 03 December 2013, 23:36:33 UTC |
99457fd | Dan LaRocque | 24 November 2013, 10:10:41 UTC | [maven-release-plugin] prepare for next development iteration | 24 November 2013, 10:10:41 UTC |
2c010b7 | Dan LaRocque | 24 November 2013, 10:10:41 UTC | [maven-release-plugin] prepare release 0.4.1 | 24 November 2013, 10:10:41 UTC |
bc068e7 | Dan LaRocque | 24 November 2013, 10:08:07 UTC | Remove Faunus mention from rewrite-tag.sh I added new error-checking to this script during the last Faunus release and copied my changes back to Titan afterward. I forgot to change a string containing the capitalized name of the project back from Faunus to Titan when copying. | 24 November 2013, 10:08:07 UTC |
f14e8a4 | Dan LaRocque | 24 November 2013, 09:57:44 UTC | Update doc commit pointer | 24 November 2013, 09:57:44 UTC |
8f03cb8 | Dan LaRocque | 24 November 2013, 09:57:17 UTC | Remove hazelcast mentions from website index | 24 November 2013, 09:57:17 UTC |
377ffbb | Dan LaRocque | 24 November 2013, 09:36:36 UTC | Improve titan.sh's help/usage message | 24 November 2013, 09:36:36 UTC |
bac1978 | Dan LaRocque | 24 November 2013, 05:14:08 UTC | Tweaking JRE6 pom stylesheet | 24 November 2013, 05:14:08 UTC |
6bb59f3 | Dan LaRocque | 24 November 2013, 05:08:09 UTC | Avoid overlapping template matches in jre6.xslt When multiple template patterns in a stylesheet match, Saxon throws a warning and continues parsing. In this case the multiple match is harmless, but the warning is scary so I'm making a minor change to avoid the match overlap. | 24 November 2013, 05:08:09 UTC |
6709fe2 | Dan LaRocque | 24 November 2013, 05:07:01 UTC | pom whitespace change | 24 November 2013, 05:07:01 UTC |
0d7ed3a | Dan LaRocque | 24 November 2013, 05:05:21 UTC | pom.xml: move Saxon version declaration It seems that plugin configurations don't care about dependencyManagement versions. I would probably have to use pluginManagement here, but that just makes it less readable since the plugin is only used one time. | 24 November 2013, 05:06:17 UTC |
0878f4e | Dan LaRocque | 24 November 2013, 04:48:43 UTC | Reduced loglevel of needlessly severe test log msg There was no reason for these two messages to be logged at ERROR. They belong at DEBUG or at most INFO. | 24 November 2013, 04:48:43 UTC |
218a912 | Dan LaRocque | 24 November 2013, 04:48:01 UTC | Removed ES from titan-dist-cassandra Moved Cassandra + ES tests into titan-dist-all and made titan-dist-server include those new tests. Also added a BDB + ES test in titan-dist-all. | 24 November 2013, 04:48:01 UTC |
228342f | Dan LaRocque | 24 November 2013, 04:05:08 UTC | Remove stale reference to jre7.xslt I removed this stylesheet a few commits ago, but a reference to it lived on in pom.xml. Deleted. | 24 November 2013, 04:05:08 UTC |
f32ae17 | Dan LaRocque | 24 November 2013, 04:03:30 UTC | Remove jre7 profiles from poms JRE7-compatible behavior is now the pom default. Keeping these profiles no longer makes sense now that I've added a stylesheet to rewrite them for JRE6 builds, and Maven profile activation is error prone anyway (e.g. activating one or more profiles on the command line deactivates all profiles that have <activeByDefault>true</activeByDefault>). | 24 November 2013, 04:03:30 UTC |
0eda1f3 | Dan LaRocque | 24 November 2013, 02:23:45 UTC | Set 0.4.1's scheduled release date | 24 November 2013, 02:23:45 UTC |
2f1617f | Dan LaRocque | 23 November 2013, 12:27:09 UTC | Fixing titan-dist-all, -server jre6 packaging bug The jre6.xslt stylesheet now removes the persistit dependencies from these two distro poms. I also deleted the jre7.xslt stylesheet, since reversing this operation is nontrivial. So now `git checkout` is the way to reverse it. | 23 November 2013, 12:27:09 UTC |
1abf2ed | Dan LaRocque | 23 November 2013, 11:09:00 UTC | Removing uses of Java 7's ProcessBuilder.Redirect This is a convenience introduced in Java SE 7. I recently used it in HBase's test harness and in the distribution failsafe tests without realizing that it wasn't present in Java 6. This commit replaces it with a clunkier but functionally equivalent Java 6 alternative. | 23 November 2013, 11:09:00 UTC |
0458891 | Dan LaRocque | 23 November 2013, 11:07:40 UTC | Flesh out JRE6/7 targeting XSLT sheets Added retargeting support to the distribution archives, including both their contents and the filenames of the zip/tar.bz2/asc files. | 23 November 2013, 11:07:40 UTC |
830a90d | Dan LaRocque | 23 November 2013, 04:21:10 UTC | Adding pom XSLT stylesheets for JRE retargeting | 23 November 2013, 04:21:10 UTC |
d80815e | Dan LaRocque | 23 November 2013, 03:35:38 UTC | Add XSLT pom stylesheets to support JRE6 builds | 23 November 2013, 03:35:38 UTC |
b9f5090 | Dan LaRocque | 22 November 2013, 16:22:21 UTC | Remove obsolete, unused titan-serve conf file titan-server-rexster.xml is no longer needed now that RexsterTitanServer is out | 22 November 2013, 16:22:21 UTC |
b4940b3 | Matthias Broecheler | 21 November 2013, 03:36:36 UTC | fixed eventual graph test to reflect recent updates | 21 November 2013, 03:36:36 UTC |
c97601a | Matthias Broecheler | 21 November 2013, 02:43:09 UTC | Added metrics test cases for Cassandra and Hbase. Fixed bug that circumvented type caching (related to #489). | 21 November 2013, 02:43:09 UTC |
6576a7a | Matthias Broecheler | 21 November 2013, 02:11:45 UTC | Set cache concurrency to available processors | 21 November 2013, 02:11:45 UTC |
06755bf | Matthias Broecheler | 19 November 2013, 22:56:38 UTC | Fixed exception message | 19 November 2013, 22:56:38 UTC |
ca48986 | Matthias Broecheler | 19 November 2013, 22:21:00 UTC | use Guava cache in RelationQueryCache | 19 November 2013, 22:21:00 UTC |
674dd4c | Matthias Broecheler | 19 November 2013, 09:35:15 UTC | computed size of guava entry with soft values | 19 November 2013, 09:35:15 UTC |
a145711 | Matthias Broecheler | 19 November 2013, 08:46:28 UTC | updated changelog. | 19 November 2013, 08:46:28 UTC |
19d20cb | Matthias Broecheler | 19 November 2013, 08:11:42 UTC | Removed streaming of large adjacency result sets since such iterative updating of limits is expensive and has very little benefit. Users should set a limit up front. In the REPL, a CTRL-C type functionality should be available to abort long running queries. | 19 November 2013, 08:11:42 UTC |
a1870b3 | Matthias Broecheler | 19 November 2013, 07:59:31 UTC | improved database level cache sizing | 19 November 2013, 07:59:31 UTC |
0bcede8 | Matthias Broecheler | 19 November 2013, 07:10:51 UTC | updated changelog, simplifed parameter passing. | 19 November 2013, 07:10:51 UTC |
ee21e37 | Matthias Broecheler | 19 November 2013, 04:28:11 UTC | updated cache configuration for non-embedded local configuration to use 3 minute timeout. | 19 November 2013, 04:28:11 UTC |
1a19b02 | Matthias Broecheler | 19 November 2013, 04:20:02 UTC | Updated configuration files to enable cache by default | 19 November 2013, 04:20:02 UTC |
10a0e03 | Matthias Broecheler | 19 November 2013, 04:18:54 UTC | Added a concurrent update test for db level cache. | 19 November 2013, 04:18:54 UTC |
1887ef9 | Matthias Broecheler | 19 November 2013, 01:00:40 UTC | Improved database cache test coverage | 19 November 2013, 01:00:40 UTC |
7311776 | Matthias Broecheler | 19 November 2013, 00:50:53 UTC | Better size estimation for cache | 19 November 2013, 00:50:53 UTC |
9c539d8 | Matthias Broecheler | 18 November 2013, 23:04:00 UTC | needed to change order of tests for results to line up. | 18 November 2013, 23:04:00 UTC |
19a94c8 | Matthias Broecheler | 18 November 2013, 22:53:35 UTC | Added additional counting test case | 18 November 2013, 22:53:35 UTC |
aa5049e | Matthias Broecheler | 18 November 2013, 22:50:42 UTC | Added additional counting test case | 18 November 2013, 22:50:42 UTC |
7bfdab7 | Matthias Broecheler | 18 November 2013, 21:56:36 UTC | Updated changelog, upgrade, readme and notice for the upcoming 0.4.1 release | 18 November 2013, 21:56:36 UTC |
6807178 | Matthias Broecheler | 18 November 2013, 21:56:17 UTC | Updated indexing predicates to distinguish between token based and entire string based PREFIX and REGEX queries. Implemented those changes throughout the stack and updated test cases. | 18 November 2013, 21:56:17 UTC |
b4ade65 | Dan LaRocque | 17 November 2013, 14:49:57 UTC | Removing ${titan.classifier} from poms | 17 November 2013, 14:49:57 UTC |
f8404ce | Dan LaRocque | 17 November 2013, 14:42:55 UTC | Cygwin compatibility for titan-server | 17 November 2013, 14:42:55 UTC |
a146d98 | Dan LaRocque | 16 November 2013, 09:29:50 UTC | Further reduced testTransactionalMemory iterations | 16 November 2013, 09:29:50 UTC |
d5e9a62 | Dan LaRocque | 16 November 2013, 08:53:25 UTC | Reducing iteration count in testTranactionalMemory A single execution of this method takes approximately 35 minutes on an EC2 m1.medium. This is unnecessarily for performance testing purposes with JUnitBenchmarks. I reduced the iteration count by an order of magnitude (100k to 10k) and made the thread count equal to the number of cores * 2 rather than the hardcoded number 8. | 16 November 2013, 08:53:25 UTC |
eb38250 | Dan LaRocque | 16 November 2013, 06:13:17 UTC | Tweaks to performance test surefire config | 16 November 2013, 06:13:17 UTC |
96913c0 | Dan LaRocque | 16 November 2013, 06:10:52 UTC | Make HBase shutdown more carefully in tests When running memory and performance tests through Maven, I found that HBase would die after perf tests but fail to restart for mem tests. This turned out to be ZK failing to bind its port. The HBase test harness killed the HBase process with SIGTERM, which seems to prevent ZK from cleanly closing its listen socket, which then goes into TIME_WAIT and prevents the next ZK process from binding the port. To fix this, I made the test harness first try to stop HBase using the clean shutdown script provided with the HBase distro, waiting up to 20 seconds for it to finish. If it doesn't finish, the harness resorts to SIGKILL (not TERM). This approach seems to resolve the issue -- now perf and mem tests both run successfully. | 16 November 2013, 06:10:52 UTC |
d4974bc | Dan LaRocque | 16 November 2013, 03:55:19 UTC | Adding two new Cassandra memory test subclasses | 16 November 2013, 03:55:19 UTC |
156c92d | Dan LaRocque | 16 November 2013, 03:54:14 UTC | Increasing memory in Cassandra + HBase tests | 16 November 2013, 03:54:14 UTC |
372e190 | Dan LaRocque | 16 November 2013, 03:52:39 UTC | Adding JUnitBenchmark env var config options | 16 November 2013, 03:52:39 UTC |
5f09013 | Dan LaRocque | 16 November 2013, 03:52:16 UTC | Deleting chronically unmaintained External* tests | 16 November 2013, 03:52:16 UTC |
5aedfbe | Matthias Broecheler | 15 November 2013, 22:04:41 UTC | removed infinispan from titan-dist | 15 November 2013, 22:04:41 UTC |
211df93 | Matthias Broecheler | 15 November 2013, 21:59:21 UTC | removed infinispan from titan-all | 15 November 2013, 21:59:21 UTC |
9b59abf | Matthias Broecheler | 15 November 2013, 21:08:19 UTC | Reverted back to Guava cache which is slightly slower but better behaved since it does not require a separate cleanup thread. | 15 November 2013, 21:08:19 UTC |