Totally unacceptable non-performance 21 million records in file geodatabase table Selecting by a single attribute ADDRESS_TYPE = 'A' via Make Feature Layer. Deleting with Delete Features outside of edit session. Eerily, the Make Feature Layer takes seconds - did it fail? The attribute has an attribute index built on it. 18% complete is 2.5 hours.
I manually selected the same records from the attribute table in a different copy of the data. Then ran Delete Features outside an edit session. Performance similarly slow.
What can be done?
Suggestions welcome. Simply cannot do my job with 10.0 without speed up. Previously such operations were not impossibly slow.
Totally unacceptable non-performance 21 million records in file geodatabase table Selecting by a single attribute ADDRESS_TYPE = 'A' via Make Feature Layer. Deleting with Delete Features outside of edit session. Eerily, the Make Feature Layer takes seconds - did it fail? The attribute has an attribute index built on it. 18% complete is 2.5 hours.
I manually selected the same records from the attribute table in a different copy of the data. Then ran Delete Features outside an edit session. Performance similarly slow.
What can be done?
Suggestions welcome. Simply cannot do my job with 10.0 without speed up. Previously such operations were not impossibly slow.
Actually, when I returned to work on Monday, the delete was 70% complete. That means it tooks days to delete only 70% of the records. Killed the process. Going to back ups.
You will be better off making the inverse selection and exporting the records to a new feature class than deleting the selected records. At the 10.1 release there is a new tool for (fast) table truncation which will help with these workflows.
Thanks, Bruce. I ended up doing that when it failed the first time after running a weekend. It worked well, and I did have space for the extra temporary table. I look forward to the 10.1 table operation speedups.
