Export hangs in 11.2.0.1

Question

We recently upgraded our databases (in Dec) from 9.2.0.6 to 11.2.0.1. After the upgrade, the export job (using original exp)hangs intermittently. It runs fine for a few days and then hangs. The export usually finishes in 1-3 minutes. But, when it hangs, the database needs to be bounced to kill the process. Also, a shutdown abort is required as shutdown immediate also hangs. The alert log doesn't show any errors. I have a ticket open with Oracle for this and they too do not know the cause and are still researching it. The hang is reproducible. After a bounce, the export works fine for 3-4 times and then on the 5th try. The platform is Solaris 10.
It always hangs on " exporting system procedural objects and actions " and the sql that it hangs on is -
SELECT DISTINCT METRIC_ID,
WARNING_OPERATOR,
WARNING_VALUE,
CRITICAL_OPERATOR,
CRITICAL_VALUE,
OBSERVATION_PERIOD,
CONSECUTIVE_OCCURRENCES,
INSTANCE_NAME,
OBJECT_TYPE,
OBJECT_NAME
FROM DBA_THRESHOLDS, V$METRICNAME
WHERE STATUS = 'VALID'
AND OBJECT_TYPE = 'SYSTEM'
AND GROUP_ID IN (2, 3)
AND METRICS_NAME = METRIC_NAME
UNION ALL
SELECT METRIC_ID,
WARNING_OPERATOR,
WARNING_VALUE,
CRITICAL_OPERATOR,
CRITICAL_VALUE,
OBSERVATION_PERIOD,
CONSECUTIVE_OCCURRENCES,
INSTANCE_NAME,
OBJECT_TYPE,
OBJECT_NAME
FROM DBA_THRESHOLDS, V$METRICNAME
WHERE STATUS = 'VALID'
AND OBJECT_TYPE = 'FILE'
AND GROUP_ID = 7
AND METRICS_NAME = METRIC_NAME
UNION ALL
SELECT DISTINCT METRIC_ID,
WARNING_OPERATOR,
WARNING_VALUE,
CRITICAL_OPERATOR,
CRITICAL_VALUE,
OBSERVATION_PERIOD,
CONSECUTIVE_OCCURRENCES,
INSTANCE_NAME,
OBJECT_TYPE,
OBJECT_NAME
FROM DBA_THRESHOLDS, V$METRICNAME
WHERE STATUS = 'VALID'
AND OBJECT_TYPE = 'SERVICE'
AND GROUP_ID IN (6, 10)
AND METRICS_NAME = METRIC_NAME
UNION ALL
SELECT METRIC_ID,
WARNING_OPERATOR,
WARNING_VALUE,
CRITICAL_OPERATOR,
CRITICAL_VALUE,
OBSERVATION_PERIOD,
CONSECUTIVE_OCCURRENCES,
INSTANCE_NAME,
OBJECT_TYPE,
OBJECT_NAME
FROM DBA_THRESHOLDS, V$METRICNAME
WHERE STATUS = 'VALID'
AND OBJECT_TYPE = 'EVENT_CLASS'
AND GROUP_ID = 1
AND METRICS_NAME = METRIC_NAME
UNION ALL
SELECT METRIC_ID,
WARNING_OPERATOR,
WARNING_VALUE,
CRITICAL_OPERATOR,
CRITICAL_VALUE,
OBSERVATION_PERIOD,
CONSECUTIVE_OCCURRENCES,
INSTANCE_NAME,
OBJECT_TYPE,
OBJECT_NAME
FROM DBA_THRESHOLDS, V$METRICNAME
WHERE STATUS = 'VALID'
AND OBJECT_TYPE = 'TABLESPACE'
AND WARNING_VALUE <> '<SYSTEM-GENERATED THRESHOLD>'
AND GROUP_ID = 9
AND METRICS_NAME = METRIC_NAME
This seems like a hang but after looking at the hang analysis, support says it is not a hang.. but more of a wait.
The exp command used is -
exp export/export file=/tmp/rcat.dmp full=y LOG=/opt/app/oracle/admin/backups/log/rcat_2011Jan24.log DIRECT=Y CONSTRAINTS=Y GRANTS=Y TRIGGERS=Y CONSISTENT=Y STATISTICS=NONE
The output for -
SELECT event FROM V$SESSION_WAIT WHERE SID =218 is
db file sequential read
Has anyone experienced this issue? Any ideas on what to check or how to handle this?

Answer 1

user12158503 wrote:
I am a little confused on dbms_stats.gather_dictionary_stats and dbms_stats.gather_schema_stats. The comment in the package of the dbms_stats says that dbms_stats.gather_dictionary_stats collects statistics for dictionary schemas 'SYS', 'SYSTEM' and schemas of RDBMS components. Is that the same as doing the dbms_stats.gather_schema_stats on 'SYS' and 'SYSTEM'? I don't know what the rdbms components are. I ran the dbms_stats.gather_dictionary_stats collects and see that statisticss has not been collected on many of the 'SYS' and 'SYSTEM' tables . The last_analyzed column still shows a date 3 months ago. Why is this so?
Can anyone please explain this?AFAIK that is depend of options parameter`s value of this procedure.So default that is GATHER AUTO but if you use just GATHER then this last_analyzed column column will change.