[
https://issues.jboss.org/browse/TEIID-2879?page=com.atlassian.jira.plugin...
]
Steven Hawkins commented on TEIID-2879:
---------------------------------------
There is already some minimal logic to throttle cpu, which is done via the timeslicing
property on the engine (if processing exceeds the timeslice, then the plan will be
requeued for processing). However this doesn't provide any hard guarantees and
isn't based upon any attributes of the query.
A problem we have with any throttling scheme performed on an active basis is that there
are other resources that can be held by the plan - source connections and memory (the heap
footprint of an inactive plan is typically manageable, but intermediate data spilled to
disk can be significant). We have tracking logic already to see what source connections
are held by a plan, but we don't track buffer resources to that level (only down to
the connection). So without additional tracking logic to be more reactive, we'd have
to apply throttling before the plan got started (similar to our max active plans
setting).
Active heap throttling is somewhat manageable with the recommendation from the admin
guide:
"In systems where large intermediate results are normal (scrolling cursors or sorting
over millions of rows) you can consider increasing the max-processing-kb and decreasing
the max-reserve-kb so that each request has access to an effectively smaller buffer
space."
That is tell Teiid to use a more per plan allocation scheme rather than using a more
shared memory pool.
Add more Resource Management
----------------------------
Key: TEIID-2879
URL:
https://issues.jboss.org/browse/TEIID-2879
Project: Teiid
Issue Type: Feature Request
Reporter: Rick Wagner
Assignee: Steven Hawkins
To protect the system sometimes we may need to throttle the queries to ensure it does not
exhaust all the existing resources on the server. Resource management is the key to play
this role of monitoring and throttling queries. Oracle products offer such a feature for
managing the consumption of resources. We have quite a number of (Teiid) profiles running
on a big box and we want to ensure no silly queires are hogging on to all the resources
and consequently causing issue to others.
Ability to throttle the query by restricting query that uses
1) Too much CPU
2) Too much java heap
3) Long running query
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:
http://www.atlassian.com/software/jira