By using AWS re:Post, you agree to the AWS re:Post Terms of Use
AWS re:Postre:Post

Can't understand performance differences between two similar DMS CDC tasks

0

I have two Oracle databases that I am using DMS to capture CDC changes to S3. Both are configured the same way - they are two different applications from the same vendor, and the vendor uses the same middleware/app server software for both, with common configurations for the Oracle databases. I have set up DMS using CloudFormation templates that I copied one from the other and edited - the only things that are different between the two are naming, the IP address of the Oracle server, things like that.

All of this is to say I feel justified expecting the two migration tasks to behave the same. But they don't! One goes from a restart command to reading archive logs and writing change files in about 30 minutes (there are hundreds of tables in each of these databases). The other one takes several hours. The only substantive difference I can think of is that I created all the DMS infrastructure for the fast database a couple of years before the slow one, and DMS software versions changed in between. It's a problem because we routinely make database schema changes and stop and start the Oracle databases for preventive maintenance.

What are some things I can look at to understand this difference in behavior? Thank you.

Topics
Migration & Modernization
Tags
AWS Database Migration Service
Language
English
srenker
asked 5 months ago196 views
1 Answer
0

The first thing I would check is the DMS cloudwatch logs which will clarify where the DMS task is waiting for the one which takes hours. DMS also creates some control tables on the target database which hold useful information about the replication state and status. Those tables are stored in dmslogs schema. You mentioned that the two databases support two different applications from same vendor. If those apps are different is it possible one apps is more heavily used than the other and correspondingly the underlying database for the same is more active and generate more changes than the other ? data volume can be a factor for replication lag. DMS latest version is supposed to be more performant than previous so I don't think version would be a problem unless you are hitting any bug which will show up in cloudwatch log. Anyway pls confirm the DMS version here and reply back.

AWS
rePost-User-0871803
answered 5 months ago
  • srenker
    5 months ago

    Thanks for your response. DMS version is 3.5.1. You make a good point about data volume but I observe this difference even in development systems where the data volume is extremely low for both applications.

    I have done a cursory review of logs and one difference I see is: in the faster database, the "initial load" of tables (I forgot to mention this is a CDC-only load, and I don't understand why it starts with an initial load of 0 rows for every table) has each subtask go on to the next table when it is finished with the table it is working on, whereas in the slower database it seems to wait for all subtasks to complete before going on to the next set of tables.

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions

Relevant content