[ https://issues.apache.org/jira/browse/HIVE-22402?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16960049#comment-16960049 ]
Hive QA commented on HIVE-22402: -------------------------------- Here are the results of testing the latest attachment: https://issues.apache.org/jira/secure/attachment/12984046/HIVE-22402.2.patch {color:red}ERROR:{color} -1 due to no test(s) being added or modified. {color:red}ERROR:{color} -1 due to 22 failed/errored test(s), 17545 tests executed *Failed tests:* {noformat} org.apache.hadoop.hive.cli.TestAccumuloCliDriver.testCliDriver[accumulo_predicate_pushdown] (batchId=291) org.apache.hadoop.hive.cli.TestAccumuloCliDriver.testCliDriver[accumulo_queries] (batchId=291) org.apache.hadoop.hive.cli.TestAccumuloCliDriver.testCliDriver[accumulo_single_sourced_multi_insert] (batchId=291) org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[symlink_text_input_format] (batchId=52) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[external_table_ppd] (batchId=105) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_binary_external_table_queries] (batchId=108) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_binary_map_queries] (batchId=109) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_binary_map_queries_prefix] (batchId=107) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_binary_storage_queries] (batchId=108) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_handler_snapshot] (batchId=108) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_joins] (batchId=110) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_null_first_col] (batchId=105) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_ppd_join] (batchId=107) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_ppd_key_range] (batchId=106) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_pushdown] (batchId=106) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_queries] (batchId=107) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_single_sourced_multi_insert] (batchId=109) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_timestamp] (batchId=108) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbase_viewjoins] (batchId=106) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[hbasestats] (batchId=106) org.apache.hadoop.hive.cli.TestHBaseCliDriver.testCliDriver[ppd_key_ranges] (batchId=105) org.apache.hive.service.server.TestInformationSchemaWithPrivilege.test (batchId=282) {noformat} Test results: https://builds.apache.org/job/PreCommit-HIVE-Build/19161/testReport Console output: https://builds.apache.org/job/PreCommit-HIVE-Build/19161/console Test logs: http://104.198.109.242/logs/PreCommit-HIVE-Build-19161/ Messages: {noformat} Executing org.apache.hive.ptest.execution.TestCheckPhase Executing org.apache.hive.ptest.execution.PrepPhase Executing org.apache.hive.ptest.execution.YetusPhase Executing org.apache.hive.ptest.execution.ExecutionPhase Executing org.apache.hive.ptest.execution.ReportingPhase Tests exited with: TestsFailedException: 22 tests failed {noformat} This message is automatically generated. ATTACHMENT ID: 12984046 - PreCommit-HIVE-Build > Deprecate Hive PerfLogger > ------------------------- > > Key: HIVE-22402 > URL: https://issues.apache.org/jira/browse/HIVE-22402 > Project: Hive > Issue Type: Improvement > Affects Versions: 4.0.0 > Reporter: David Mollitor > Assignee: David Mollitor > Priority: Major > Attachments: HIVE-22402.1.patch, HIVE-22402.2.patch > > > Recently I wanted to add some additional capability, and add more, > performance logging to support my troubleshooting efforts. I started looking > at PerfLogger and started to examine its usage. I discovered a few things: > # Since 'loggers' must be open and closed manually, I found a couple of > places where loggers were opened, but not closed, rendering them useless > # Since 'loggers' must be closed manually, I found a few places where an > early-return or Exception thrown would cause a logger to not be closed, > thereby rendering it useless > # Session information is not logged, so it can be difficult to precisely > pinpoint which session is taking lots of time > # PerfLogger overloaded. Most of the time, it's being used as a simple timer > mechanism with automatic logging in SLF4J debug. However, it is also a facade > over the Hive Metrics subsystem and timing results are automatically > published to Metrics and then there becomes this dependency on a 'logger' to > be able to access metric data as well. > The last bullet is the most challenging part and why I propose to deprecate > the Hive {{PerfLogger}} and not simply remove it. I am proposing a new > system... a {{PerfTimer}} that is allows for Java 8's try-with-resources > feature to protect against the developer having to care about manually close > measurements and not having to carefully consider all early-exits. The base > implementation logs to SLF4J. An extended version automatically publishes to > the Hive Metric subsystem as well. > The Hive {{PerfLogger}} has a bit of a clunky system for allowing plugable > implementations. However, there are sections of the code that rely on the > {{PerfLogger}}'s ability to publish to the Metrics subsystem. These code > section look up various timers in the Metrics Subsytem and publish the > results back to the client. Since, in theory, the implementation is plugable, > any other implementation that does not publish to the Metrics Subsystem will > break these code paths. Also, these code paths create and interact with > {{PerfLoggers in a static way, and then the publishing code pulls the data > from the {{PerfLogger}}}} (as a facade to the Metrics subsystem) in a static > way. Therefore, when I tried to replace the entire {{PerfLogger}} code, I > came across an issue because there is not (and should not) be a way to just > statically pull this information down from any point in the code. Information > that is required for publishing should be passed around within some sort of > context object, separate from the Metrics subsystem. There was no obvious way > to string a new {{PerfTimer}} to all the required locations. I propose > marking the {{PerfLogger}} as deprecated and leaving these complex section > alone. Instead, replace only the simple "I want a timer" use cases. -- This message was sent by Atlassian Jira (v8.3.4#803005)