Oracle Connector stalling with statspack process

[Neven MacEwan]

Hi - we have an Oracle connector which is configured with about 80 included tables capturing about 60k events per hour but we noticed that when the statspack process runs then the cdc events on the tables we are capturing stall for about 2 hours - are there any settings we should look at changing? Thanks.

[Chris Cranford]

Hi, 

There are a lot of interactions between statspack and the SGA depending on the level that the process is run with. Depending on the level, these interactions will cause IO contention, particularly if your hardware is already saturated. It also has the potential to flush the SGA when taking snapshots, which will directly force Oracle to hard parse a lot of data that would otherwise be looked up directly from cached buffers. This has a direct impact on LogMiner's performance, particularly when its interacting with the data dictionary.

What level is statspack executed with?

-cc

On 2/17/26 4:51 PM, 'Neven MacEwan' via debezium wrote:

Hi - we have an Oracle connector which is configured with about 80 included tables capturing about 60k events per hour but we noticed that when the statspack process runs then the cdc events on the tables we are capturing stall for about 2 hours - are there any settings we should look at changing? Thanks. --
You received this message because you are subscribed to the Google Groups "debezium" group.
To unsubscribe from this group and stop receiving emails from it, send an email to debezium+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/debezium/e18b46ff-d4b4-4248-ae4f-cc95307f1d3an%40googlegroups.com.

[Neven MacEwan]

Hi Chris

From the Oracle DBA

"To answer your questions: the statspack snapshot level that is used in FSPRDTM is 7 (slightly higher than the default level 5) and the hardware is definitely not saturated - we don't have any issues with memory/cpu utilisation on the server.

 

We don't think this comment on the Debezium support forum is relevant to the issue with CDC delays. The statspack snapshots are collected every 30min and that collection is not causing any issues for CDC. The issue occurs when the cronjob to purge the old statspack snapshot runs (it used to be at 8am, now at 1am). As discussed before, that cronjob runs just for 2-3 minutes, but it generates about 6GB of archive logs, which apparently slows down the Logminer processing, although it shouldn't even look for the changes in tables in the PERFSTAT schema that are used by Statspack. We have to run this purging cronjob, otherwise that tablespace would grow very fast. I mentioned before that we can change this cronjob to run weekly (e.g. 1am on Sundays) instead of daily, but in that case it would generate even more changes when it runs, so not sure if that would help."

Question - if we explicitly exclude the  PERFSTAT schema would that make a difference?

Neven

[Chris Cranford]

Hi Neven -

Yes, if you are not currently setting `schema.include.list`, then you can set `schema.exclude.list` and add `PERFSTAT`; however depending on your current configuration, it may not make much difference. For example, `table.include.list` already does schema filtering OOTB if you're specifying `log.mining.query.filter.mode`. What is your configuration?

-cc

To view this discussion visit https://groups.google.com/d/msgid/debezium/57e37441-c625-406d-baf3-bc639012886en%40googlegroups.com.