Stardog Performance and Timeouts

4 views
Skip to first unread message

Mark James

unread,
Jun 8, 2016, 3:35:53 AM6/8/16
to Stardog
Hi Guys,
We're in the process of migrating some queries from Virtuoso 7 to Stardog. We've found that things that run well in Virtuoso literally blow up Stardog. In Virtuoso we make extensive use of filters which don't seem to go so well in Stardog. 

I've started optimising them based on some advice in historical google groups posts, which seems to be going fairly well. I'm sure I'll have some further optimisation questions later. 

But for now the thing that I'm scratching me head about is long queries that don't seem to be killed via the timeout.

I've got the timeout set to the default 5 minutes however as can be seen below I have some very long queries running.

+----------+----------+-----------+--------------+--------------+
| Query ID | Database |   User    | Elapsed time |   Timeout    |
+----------+----------+-----------+--------------+--------------+
| 7720     | myDB     | admin     | 06:44:40.716 | 00:05:00.000 |
| 8024     | myDB     | admin     | 06:33:29.159 | 00:05:00.000 |
| 8046     | myDB     | admin     | 05:09:10.029 | 00:05:00.000 |
| 8330     | myDB     | admin     | 02:34:18.011 | 00:05:00.000 |
| 9259     | myDB     | admin     | 00:51:24.900 | 00:05:00.000 |
| 9260     | myDB     | anonymous | 00:51:24.853 | 00:05:00.000 |
+----------+----------+-----------+--------------+--------------+

I also experienced this error message at least once - Error! Java Heap Space - on one of these long running queries.
So it was definitely still doing something when it died. (And locked up the whole db in the process)

Is there something that could override the timeout?  Until I make sure I've sorted out the query performance it's important one of these queries can't bring down the whole db. 

cheers
Mark


Evren Sirin

unread,
Jun 8, 2016, 10:14:00 AM6/8/16
to Stardog
If the server is close to running out of memory then that can delay
returning answers or killing the query. However the elapsed time in
your queries cannot be explained by this delay. There might be some
messages in stardog.log that might provide some answers. What kind of
queries are you running? Is there any reasoning involved? Can you
share one or more of the queries along with the query explain output
for those queries?

Best,
Evren
> --
> -- --
> You received this message because you are subscribed to the C&P "Stardog"
> group.
> To post to this group, send email to sta...@clarkparsia.com
> To unsubscribe from this group, send email to
> stardog+u...@clarkparsia.com
> For more options, visit this group at
> http://groups.google.com/a/clarkparsia.com/group/stardog?hl=en
Reply all
Reply to author
Forward
0 new messages