Slowdown of getMap tiles due to clash with Antivirus software

[Greg Vds]

Hi,

We pushed online two weeks ago a new version of our app (https://gis.urban.brussels/brugis). This one is based on Mapstore 2021.02.02, which uses OpenLayers 5.3.0.

We observed a rather complex slowdown of layers/maps visualisation on machines inside our infrastructure. On machines outside it, everything is fine. In the end, we think it is due to a difficult to understand interaction between the way Openlayers asks the tiles and the antivirus installed in practically every machine of the IT services in the Brussels Region.

It seems that OL 5.3.0 is using a TileQueue class that sends all the getMap tiles request at the same time, in comparison to OL 4.6.4, which seems to send all the requests one after the other, and this seems to less throttle the antivirus check...

Is someone encountering the same kind of problem?

Is there a way to give a parameter to OL (maybe in the app.jsx or the localconfig.js) to change that behaviour?

We'll try to fallback on the OL 4.6.4 to avoid this "all at once getmap request", and see if this clear our problem.

Every advice on this topic most welcome :-),

Greg for BruGIS

[Greg Vds]

OL and Tile queue do not seem to be involved.

In the meantime, we discovered that Mapstore2 requires Geoserver to have the extension gs-WMTS multidimensional. Is this mandatory for the application?

Without the extension jar in the geoserver/WEB-INF/lib, we have an error 400 for each layer added (describedomains request unknown).

Once the library is added, describedomains is there in the WMTS getCapabilities, and we have LESS errors. 

Our layers with geowebcache do no more respond in error 400, but the other layers without a defined cache, they still receive an error, but this one from geowebcache (layer unknown).

So basically, Mapstore2 expects all layers to have a geowebcache tileset equivalent. Is it correct or needed?

[Greg Vds]

Again, after testing, this WMTS multidimensional extension is not the problem, as the error/warning message was present on machines with and without the antivirus.

Now we are trying to see if http/2 could not help/avoid queuing of the httprequests...