Open linghengqian opened 1 month ago
ppkarwasz
commented
1 month ago Hi @linghengqian,
Could you point us to how you generate the GraalVM image?
As almost all filters (except DenyAllFilter and Filters), MarkerFilter has a onMismatch property. I suspect that the GraalVM reflection metadata you use might lack the property, hence the problem in the native image.
linghengqian
commented
1 month ago Hi @linghengqian,
Could you point us to how you generate the GraalVM image?
As almost all filters (except
DenyAllFilterandFilters),MarkerFilterhas aonMismatchproperty. I suspect that the GraalVM reflection metadata you use might lack the property, hence the problem in the native image.
Hi, the steps I mentioned in the issue description never involve using log4j2 under GraalVM Native Image. I assume you are referring to https://github.com/apache/shardingsphere/pull/31526 . On the shardingspehre side, all nativeTest under GraalVM Native Image have excluded log4j2 in maven pom, and I am actually waiting for the release of log4j2 2.24.0.
On the hive side, the Error Log of ERROR MarkerFilter contains an invalid element or attribute "onMismatch" is obtained by executing unit tests under Hotspot JDK 8. Hive can still only be compiled with openjdk 8, and it is too early to discuss graalvm native image in the hive community.
ppkarwasz
commented
1 month ago @linghengqian,
The problem is due to a typo in your configuration file:
logger.CBORuleLogger.filter.marker.type = MarkerFilter
logger.CBORuleLogger.filter.marker.marker = FULL_PLAN
logger.CBORuleLogger.filter.marker.onMatch = DENY
logger.CBORuleLogger.filter.marker.onMisMatch = NEUTRALYou have onMisMatch instead of onMismatch. The misleading error message is due to the following code in PropertiesConfigurationBuilder:
Since you don't have an onMismatch attribute, the code creates a synthetic onMismatch attribute with an invalid null value.
Remark: Fixing this also fixes LOG4J2-2931, so I am redirecting that issue here.
PS: I noticed that Apache Hive uses a pre-2.6 properties configuration format (the loggers is not required since version 2.6).
Would you consider switching to a more easily readable configuration format? Version 3.x will introduce breaking changes to the properties format, but not the hierarchical ones.
linghengqian
commented
1 month ago @ppkarwasz
Late reply, because I need to wait for verification on apache/hive's CI side. apache/hive's CI resources are too tight, and apache/hive's jenkins node limits each contributor to 5 hours between 2 consecutive CIs. To be honest, https://github.com/apache/hive/pull/5375 is the first PR I opened on apache/hive's side, and I know almost nothing about apache/hive's logging design.
I updated https://github.com/apache/hive/pull/5375 not long ago, I removed the packages, appenders, loggers nodes of /hive/data/conf/hive-log4j2.properties, and changed logger.CBORuleLogger.filter.marker.onMisMatch to logger.CBORuleLogger.filter.marker.onMismatch. apache/hive's CI is now back on apache/logging-log4j2:2.23.0. I'm assuming this is a reasonable way forward to not break compatibility with future versions of log4j2. I'm assuming the changes are in line with what's documented in https://github.com/apache/pulsar/pull/22327 , https://logging.apache.org/log4j/2.x/manual/configuration.html , and https://logging.apache.org/log4j/2.x/manual/plugins.html . I note from https://issues.apache.org/jira/browse/LOG4J2-3644 that the packages property seems to be straightforward to remove, though https://github.com/apache/jmeter/issues/5937 on the apache/jmeter side is still unclear as to the way forward.
2024-08-02T10:55:16.634Z main WARN The use of package scanning to locate plugins is deprecated and will be removed in a future release
2024-08-02T10:55:16.650Z main WARN The use of package scanning to locate plugins is deprecated and will be removed in a future release
2024-08-02T10:55:16.670Z main WARN The use of package scanning to locate plugins is deprecated and will be removed in a future release
2024-08-02T10:55:16.686Z main WARN The use of package scanning to locate plugins is deprecated and will be removed in a future releaseOn the apache/hive side, I noticed a new issue when bump from apache/logging-log4j2:2.23.0 to apache/logging-log4j2:2.23.1, the status = INFO defined in hive-log4j2.properties is no longer valid, and Log4j2 internal logs at TRACE and DEBUG levels will be printed starting from apache/logging-log4j2:2.23.1. Since the current issue is bound to https://issues.apache.org/jira/browse/LOG4J2-2931 , I opened #2794 to avoid ambiguity.
I will say that I personally don't know how to fix #2791 with the good first issue tag.
ppkarwasz
commented
1 month ago @ppkarwasz
I note from https://issues.apache.org/jira/browse/LOG4J2-3644 that the
packagesproperty seems to be straightforward to remove, though Migrate log4j2 to logback jmeter#5937 on theapache/jmeterside is still unclear as to the way forward.
There is no problem in Apache Hive: I checked that the appropriate annotation processor is configured in the hive-exec project and there is a Log4j2Plugins.dat file in the JAR. You can safely remove the packages configuration attribute.
* I will say that I personally don't know how to fix [Fix handling of `onMatch` and `onMismatch` attributes in the properties configuration format #2791](https://github.com/apache/logging-log4j2/issues/2791) with the `good first issue` tag.
I believe that the problem breaks down to a missing null-check: ComponentBuilder#addAttribute should add an attribute if the value is not null and remove it if the value is null.
null is never a valid value for an attribute.
BTW: All the problems you are having are due to the fact that TestHplsqlLocal and TestHplsqlOffline are fragile tests. They write something to System.out and they make assertions about that. The problem is: other components (like the status logger) might also write to System.out. I proposed a solution to this in apache/hive#5381
linghengqian
commented
1 month ago Hi @linghengqian,
Could you point us to how you generate the GraalVM image?
As almost all filters (except
DenyAllFilterandFilters),MarkerFilterhas aonMismatchproperty. I suspect that the GraalVM reflection metadata you use might lack the property, hence the problem in the native image.
Description
onMismatchofMarkerFilteris no longer available. This was actually found from https://github.com/apache/hive/pull/5375 . It can be reproduced in both2.23.0and2.23.1. This will be reflected in a log similar toERROR MarkerFilter contains an invalid element or attribute "onMismatch".Configuration
Version: [Log4j version]
2.23.0Operating system: [OS and version]
Ubuntu WSL 22.04.4JDK: [JDK distribution and version]
8.0.422-temonSDKMAN!Logs
Reproduction
/hive/data/conf/hive-log4j2.properties.property.hive.log.level = DEBUG property.hive.root.logger = DRFA property.hive.log.dir = ${sys:test.tmp.dir}/log property.hive.log.file = hive.log property.hive.test.console.log.level = INFO
appenders = console, DRFA
appender.console.type = Console appender.console.name = console appender.console.target = SYSTEM_ERR appender.console.layout.type = PatternLayout appender.console.layout.pattern = %d{ISO8601} %5p [%t] %c{2}: %m%n
appender.DRFA.type = RollingRandomAccessFile appender.DRFA.name = DRFA appender.DRFA.fileName = ${sys:hive.log.dir}/${sys:hive.log.file} appender.DRFA.filePattern = ${sys:hive.log.dir}/${sys:hive.log.file}.%d{yyyy-MM-dd} appender.DRFA.layout.type = PatternLayout appender.DRFA.layout.pattern = %d{ISO8601} %5p [%t] %c{2}: %m%n appender.DRFA.policies.type = Policies appender.DRFA.policies.time.type = TimeBasedTriggeringPolicy appender.DRFA.policies.time.interval = 1 appender.DRFA.policies.time.modulate = true appender.DRFA.strategy.type = DefaultRolloverStrategy appender.DRFA.strategy.max = 30
loggers = HadoopIPC, HadoopSecurity, Hdfs, HdfsServer, HadoopMetrics2, Mortbay, Yarn, YarnServer, Tez, HadoopConf, Zookeeper, ServerCnxn, NIOServerCnxn, ClientCnxn, ClientCnxnSocket, ClientCnxnSocketNIO, DataNucleus, Datastore, JPOX, Operator, Serde2Lazy, ObjectStore, CalcitePlanner, AmazonAws, ApacheHttp, Thrift, Jetty, BlockStateChange, swo, CBORuleLogger
logger.HadoopIPC.name = org.apache.hadoop.ipc logger.HadoopIPC.level = WARN
logger.HadoopSecurity.name = org.apache.hadoop.security logger.HadoopSecurity.level = INFO
logger.Hdfs.name = org.apache.hadoop.hdfs logger.Hdfs.level = INFO
logger.HdfsServer.name = org.apache.hadoop.hdfs.server logger.HdfsServer.level = WARN
logger.HadoopMetrics2.name = org.apache.hadoop.metrics2 logger.HadoopMetrics2.level = INFO
logger.Mortbay.name = org.mortbay logger.Mortbay.level = INFO
logger.Yarn.name = org.apache.hadoop.yarn logger.Yarn.level = INFO
logger.YarnServer.name = org.apache.hadoop.yarn.server logger.YarnServer.level = WARN
logger.Tez.name = org.apache.tez logger.Tez.level = INFO
logger.HadoopConf.name = org.apache.hadoop.conf.Configuration logger.HadoopConf.level = ERROR
logger.Zookeeper.name = org.apache.zookeeper logger.Zookeeper.level = INFO
logger.ServerCnxn.name = org.apache.zookeeper.server.ServerCnxn logger.ServerCnxn.level = WARN
logger.NIOServerCnxn.name = org.apache.zookeeper.server.NIOServerCnxn logger.NIOServerCnxn.level = WARN
logger.ClientCnxn.name = org.apache.zookeeper.ClientCnxn logger.ClientCnxn.level = WARN
logger.ClientCnxnSocket.name = org.apache.zookeeper.ClientCnxnSocket logger.ClientCnxnSocket.level = WARN
logger.ClientCnxnSocketNIO.name = org.apache.zookeeper.ClientCnxnSocketNIO logger.ClientCnxnSocketNIO.level = WARN
logger.DataNucleus.name = DataNucleus logger.DataNucleus.level = ERROR
logger.Datastore.name = Datastore logger.Datastore.level = ERROR
logger.JPOX.name = JPOX logger.JPOX.level = ERROR
logger.Operator.name = org.apache.hadoop.hive.ql.exec.Operator logger.Operator.level = INFO
logger.Serde2Lazy.name = org.apache.hadoop.hive.serde2.lazy logger.Serde2Lazy.level = INFO
logger.ObjectStore.name = org.apache.hadoop.hive.metastore.ObjectStore logger.ObjectStore.level = INFO
logger.CalcitePlanner.name = org.apache.calcite.plan.RelOptPlanner logger.CalcitePlanner.level = INFO
logger.CBORuleLogger.name = org.apache.hadoop.hive.ql.optimizer.calcite.RuleEventLogger logger.CBORuleLogger.level = OFF logger.CBORuleLogger.filter.marker.type = MarkerFilter logger.CBORuleLogger.filter.marker.marker = FULL_PLAN
logger.CBORuleLogger.filter.marker.onMatch = DENY logger.CBORuleLogger.filter.marker.onMisMatch = NEUTRAL
logger.AmazonAws.name=com.amazonaws logger.AmazonAws.level = INFO
logger.ApacheHttp.name=org.apache.http logger.ApacheHttp.level = INFO
logger.Thrift.name = org.apache.thrift logger.Thrift.level = WARN
logger.Jetty.name = org.eclipse.jetty logger.Jetty.level = WARN
logger.BlockStateChange.name = BlockStateChange logger.BlockStateChange.level = WARN
rootLogger.level = ${sys:hive.log.level} rootLogger.appenderRefs = root, console rootLogger.appenderRef.root.ref = ${sys:hive.root.logger} rootLogger.appenderRef.console.ref = console rootLogger.appenderRef.console.level = ${sys:hive.test.console.log.level}
logger.swo.name = org.apache.hadoop.hive.ql.optimizer.SharedWorkOptimizer logger.swo.level = DEBUG