Hi,
i have a impdp job that did run for a long time without error. (OS= Solaris 11.1 DB=11.2.0.3 RAC)
impdp user/pwd schemas=XXX directory=data_pump_dir logfile=XXX.log network_link=DBL_XXX parallel=4 cluster=n flashback_time=SYSTIMESTAMP
Now i receive following error:
ORA-39001: invalid argument value
ORA-39150: bad flashback time
ORA-08186: invalid timestamp specified
ORA-06512: at "SYS.TIMESTAMP_TO_SCN", line 1
I did a run with datapump trace enabled and the logfile shows this:
KUPM:08:30:27.204: FLASHBACK_TIME expression is: SELECT TO_CHAR(SYSTIMESTAMP, 'YY-MON-DD HH:MI:SS AM') FROM SYS.DUAL
KUPM:08:30:27.204: value_t after expression conversion: 14-APR-10 08:30:27 AM
KUPM:08:30:27.205: Evaluated FLASHBACK_TIME is: 14-APR-10 08:30:27 AM
KUPM:08:30:27.205: Timestamp to scn call is: SELECT TIMESTAMP_TO_SCN@DBL_ADICOMP(TO_TIMESTAMP(:1, :2)) FROM SYS.DUAL
KUPM:08:30:27.225: ORA-39150: bad flashback time
KUPM:08:30:27.226: ORA-08186: invalid timestamp specified
ORA-06512: at "SYS.TIMESTAMP_TO_SCN", line 1
This started to happen after Daylight saving change in Germany on 30 March. Up until that point this job was running fine. I have read all related notes to nls, daylight saving, etc and checked everything i could think of, but without resolving this issue.
OS is Solaris and DB environment is set to TZ=Europe/Berlin. Daylight saving was handled by solaris and this did work fine. Time is fine on all shells, db connections, listeners, ...
NLS Settings are the same for all databases including the db accessed via the network link (runs on the same datacenter and timezone).
Any ideas on what this could be and where to look next?
Regards
Thomas