[mapserver-dev] Mapserver thread safety

[Michel.Dastous at bentley.com]

About WFS/WMS thread safety, looking at mapows.c I found a single static variable (epsgCode) used in msOWSGetEPSGProj().

Instead of returning a static pointer, the method could easily take a buffer + buffer size as parameters... Not a big deal except that 62 calls to that function will have to be modified.

Is this the only thing that make WFS/WFS server non thread safe?

Thanks !


From: Michel Dastous
Sent: Tuesday, August 13, 2013 9:45 AM
To: 'Umberto Nicoletti'
Cc: MapServer Dev Mailing List
Subject: RE: [mapserver-dev] Mapserver thread safety

Thank you very much! I will test it as soon as possible.

>From the page you sent me I realized that WMS/WFS server is not thread safe ("WMS/WFS server: static data used for state of dispatcher").   Is there any plan to fix this ? I could not find any ticket about this

Thanks.

From: Umberto Nicoletti [mailto:umberto.nicoletti at gmail.com]
Sent: Tuesday, August 13, 2013 8:49 AM
To: Michel Dastous
Cc: MapServer Dev Mailing List
Subject: Re: [mapserver-dev] Mapserver thread safety

FYI, I've opened this pull request which should fix the Oracle spatial issue you are seeing:
https://github.com/mapserver/mapserver/pull/4729
Let me know if it works for you.

BR,
Umberto
On Tue, Aug 13, 2013 at 8:23 AM, Umberto Nicoletti <umberto.nicoletti at gmail.com<mailto:umberto.nicoletti at gmail.com>> wrote:
Michel,
this page attempts to keep track of which Mapserver components are thread safe:

http://mapserver.org/faq.html#is-mapserver-thread-safe

it must not be completely up to date as, as you found out, Oracle is listed as safe, while it's not.
As for fixing it, it should be relatively easy as those two variables can be made local to the functions using them.

Please open an issue and I will look into it asap.

Umberto


On Mon, Aug 12, 2013 at 9:44 PM, <Michel.Dastous at bentley.com<mailto:Michel.Dastous at bentley.com>> wrote:
Hi all,

I've been testing mapserver and its c# mapscript API for a while now, and I have noticed that making simultaneous Draw request on multiple maps containing Oracle layer will end up in memory corruption.  No surprise here since maporaclespatial.c uses static variables (ordinates_tdo and ordinates).

My question is: Does anyone have already evaluated how much work it would require to make it thread safe (i.e remove those static variables from the code) ?

Furthermore, I have also tested the MSSQL plugin which is stable under same conditions, unfortunately it doesn't seem to scale very well my 4 core machines with maximum CPU usage of 15%.

I don't think the problem come from mapserver' s rendering engine... when drawing a map with inlined data, I could easily reach a 90% usage.

So any suggestions/information would be welcome.

Thanks.


_______________________________________________
mapserver-dev mailing list
mapserver-dev at lists.osgeo.org<mailto:mapserver-dev at lists.osgeo.org>
http://lists.osgeo.org/mailman/listinfo/mapserver-dev


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/mapserver-dev/attachments/20130813/32f3d3b7/attachment.html>