HDDS-4442. Make all Log4J2 loggers asynchronous

[adoroszlai]

What changes were proposed in this pull request?

Make all Log4J2 loggers asynchronous (only used for audit logging in Ozone) via the system property log4j2.contextSelector. This prevents lookup of location information by default, and "will give the best performance" by avoiding some intermediate threads.

https://issues.apache.org/jira/browse/HDDS-4442

How was this patch tested?

Updated TestOzoneAuditLogger unit test:

• include class name in the log message
• verify that only ? is logged, due to lack of location information

Temporarily made the same change in OM audit log pattern:

diff --git hadoop-ozone/dist/src/shell/conf/om-audit-log4j2.properties hadoop-ozone/dist/src/shell/conf/om-audit-log4j2.properties
index 57577e162..170d6840e 100644
--- hadoop-ozone/dist/src/shell/conf/om-audit-log4j2.properties
+++ hadoop-ozone/dist/src/shell/conf/om-audit-log4j2.properties
@@ -71,7 +71,7 @@ appender.rolling.name=RollingFile
 appender.rolling.fileName =${sys:hadoop.log.dir}/om-audit-${hostName}.log
 appender.rolling.filePattern=${sys:hadoop.log.dir}/om-audit-${hostName}-%d{yyyy-MM-dd-HH-mm-ss}-%i.log.gz
 appender.rolling.layout.type=PatternLayout
-appender.rolling.layout.pattern=%d{DEFAULT} | %-5level | %c{1} | %msg | %throwable{3} %n
+appender.rolling.layout.pattern=%d{DEFAULT} | %-5level | %c{1} | %C | %msg | %throwable{3} %n
 appender.rolling.policies.type=Policies
 appender.rolling.policies.time.type=TimeBasedTriggeringPolicy
 appender.rolling.policies.time.interval=86400

and checked OM audit log:

$ docker-compose exec om bash -c 'cat /var/log/hadoop/*audit*log'
2020-11-10 13:49:53,322 | INFO  | OMAudit | ? | user=hadoop | ip=192.168.96.6 | op=CREATE_VOLUME {admin=hadoop, owner=hadoop, volume=vol1, creationTime=1605016193288, modificationTime=1605016193288, quotaInBytes=-1, quotaInCounts=-1, objectID=256, updateID=1, usedBytes=0} | ret=SUCCESS |
2020-11-10 13:49:53,390 | INFO  | OMAudit | ? | user=hadoop | ip=192.168.96.6 | op=CREATE_BUCKET {volume=vol1, bucket=bucket1, gdprEnabled=null, acls=[user:hadoop:a[ACCESS], group:users:a[ACCESS]], isVersionEnabled=false, storageType=DISK, creationTime=1605016193368, bucketEncryptionKey=null, modificationTime=1605016193368, usedBytes=0} | ret=SUCCESS |
2020-11-10 13:49:53,587 | INFO  | OMAudit | ? | user=hadoop | ip=192.168.96.6 | op=ALLOCATE_KEY {volume=vol1, bucket=bucket1, key=vm7ozunjvq/0, dataSize=10240, replicationType=RATIS, replicationFactor=THREE} | ret=SUCCESS |
2020-11-10 13:49:57,490 | INFO  | OMAudit | ? | user=hadoop | ip=192.168.96.6 | op=COMMIT_KEY {volume=vol1, bucket=bucket1, key=vm7ozunjvq/0, dataSize=10240, replicationType=RATIS, replicationFactor=ONE} | ret=SUCCESS |

[jojochuang]

We should update the PR title...

[elek]

+1, thanks the patch @adoroszlai and the review @jojochuang

[adoroszlai]

Thanks @jojochuang for reporting and reviewing this, and @elek for reviewing and committing the fix.