How to monitor materialized views memory usage?

[m.ardito]

About matviews, docs say "Internal materialization (and temp tables in general) have memory overhead for each page. A rough guideline is that there can be 100 million rows in all materialized tables across all VDBs for every gigabyte of heap."

I have currently no apparent problem with that, but I would like to know/monitor how much memory matviews are using, to prevent excessive usage or adapt the "heap" to needed usage. Or, just to know. What/where/when should I check to get this (an maybe others) system status info in my wildfly/teiid?  And what counts here as available "heap"?

Thanks, Marco

[shawkins]

> but I would like to know/monitor how much memory matviews are using

Unfortunately we don't currently provide that.  It's also somewhat complicated given the current design.  It would be a mixture of heap estimates and actual byte storage sizes.  Specifically as it relates to page usage of heap memory, that is mostly determined by the row count.

> to prevent excessive usage or adapt the "heap" to needed usage.

The heap is only an issue for an extremely large number of pages.  The data itself is paged to disk as determined by the buffer manager settings.

> Or, just to know. What/where/when should I check to get this (an maybe others) system status info in my wildfly/teiid?

engine-statistics is exposed as an admin operation that shows a variety of stats on the buffermanager, but it doesn't specifically show the number of pages/batches held in memory.  That could be added with an enhancement request.

> And what counts here as available "heap"

Available heap would be the heap size beyond the base footprint of Teiid/WildFly and any of the other deployments.