rexpro fails on multiple queries

35 views
Skip to first unread message

Bhargav Raut

unread,
Feb 24, 2015, 2:58:20 PM2/24/15
to aureliu...@googlegroups.com
Hi,

I have been running a few tests with the rexpro java client, inside a spring app.
I have a test scenario, where 300 queries are fired at the app on a particular url using an asynchronous http request client.
I then wait for the responses.

The queries are mapped to a simple function which checks whether a particular edge exists, and if it does, a property on the edge is incremented by 1.

Results:
the first 10 or 15 queries receive responses.
I get nothing after that.

On the server side after about 150 requests are registered, rexpro hangs indefinitely. No further requests are serviced.

I need to know what I'm doing wrong.

Regards,
Bhargav.

Stephen Mallette

unread,
Feb 25, 2015, 8:16:06 AM2/25/15
to aureliu...@googlegroups.com
Have you altered your default thread pool settings in rexster.xml?  Depending on the OS (for some reason, i've never fully grasped), the default settings are not always good when there is a large number of concurrent requests.

--
You received this message because you are subscribed to the Google Groups "Aurelius" group.
To unsubscribe from this group and stop receiving emails from it, send an email to aureliusgraph...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/aureliusgraphs/04a94620-7157-47fa-965b-cf135a2a85d9%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Bhargav Raut

unread,
Feb 25, 2015, 1:19:01 PM2/25/15
to aureliu...@googlegroups.com
I have not changed anything with respect to the default settings which currently look like this
Worker
    Core size 8
    Max size 8
Kernal
    Core size 4
    Max size 4

What would you like me to change and to what limit

Thanks for your reply,
Bhargav


Sent from Samsung Mobile

Stephen Mallette

unread,
Feb 26, 2015, 9:32:58 AM2/26/15
to aureliu...@googlegroups.com
I'd encourage you to experiment a bit to see what might end up working for you (or it might not work at all, I suppose - i was just sort of guessing at your problem). how about just starting by increasing the workers by a multiple of two to see if you see any difference at all?  

Bhargav Raut

unread,
Feb 26, 2015, 11:04:29 AM2/26/15
to aureliu...@googlegroups.com
Hi stephen,
I tried it just now 
I have a 12 core intel xeon with 75 gb of ram.
Im running the titan server on it and querying remotely.
I increased by a multiple of two and then 4 and then started to see some improvements.

Given the number of cores that I have at my disposal what would be safe for production

Eg : no of cores *10 

 Regards.

Stephen Mallette

unread,
Feb 28, 2015, 9:11:48 AM2/28/15
to aureliu...@googlegroups.com
These settings seem to be strangely OS dependent so it's hard to simply say cores * X.  Consider scrolling through this issue for advice:


Reply all
Reply to author
Forward
0 new messages