oracle - Challenge in Migrating Huge Table with LOB Data -
we're trying migrate around 230 gb oracle ec2 hosted database rds. challenge db there 1 150gb table has lot of lob data. when try migrate data using oracle import/export (data pump) takes around 9.5hrs export 150 gb table , has lob data , 2hrs import dump rds whereas other tables migrated quickly. we're using instance highest configurations still don't see improvement in performance.
just see time difference exported 150gb dump again on ec2 , in second run took 3hrs time. please suggest me better approach reduce export/import time.
ps: tried use redgate tool identify schema , data differences between oracle databases tool failed perform comparison on huge lob tables.
quickest way found migrate large volumes of lobs follows:
extract
- write customer extract (java) process given range of id values.
- write non-lob data each row csv, , lob file , reference file in csv each row.
- run many extracts in parallel (these external java processes should not blocked oracle restrictions on licence)
- preferably wrote extract file system can "swung" between 2 servers. if possible have many of these fileystems have parallel extract processes running.
load
use sqlloader. has option
image_fname filler char(80), image lobfile(image_fname) terminated eof
experiment direct=y - on versions (10g) found little unreliable, , got regulars crashes in project, may have improved type of load.
again, use parallel sql loader processes possible - might want disable constraints , indexes, or use reverse indexes reduce block contention parallel loads.
you might consider partitioning on target , parallel loading each partition.
the advantage of using "swinging" file systems eliminate network bottlenecks.
these notes general guide, not specific, , require tuning , experimentation optimal mix.
Comments
Post a Comment