Hey, we run GS in a beanstalk stack (with docker) and read the content of the data directory from S3 when each instance starts as a volume mount. These instances are immutable so don't follow the same pattern you're talking about but by doing so we can manage the data directory as code in a git repo and the deployment of instances is done by our CI/CD pipeline. This approach also allows us to scale our compute horizontally as load increases.
For the record, i've found that m4.large instances with ~5gig of ram assigned to the jvm provides the best bang for buck compute wise. We run 4 of these behind an ELB at peak times and 1 during lulls. Auto scaling rules look after this for us. This is with control flow set to 8 getMap requests. We also seed our tile cache to S3 (zipped) and read it on to each instance at launch. Although this slows the launch time to ~10 minutes, it improves performance significantly after that. We played with the S3 blob store extension but our corporate proxy meant this option was not feasible for a few reasons. Happy to share more i this approach is of interest to you. -- View this message in context: http://osgeo-org.1560.x6.nabble.com/Sharing-the-GeoServer-data-folder-on-a-S3-bucket-tp5330492p5331593.html Sent from the GeoServer - User mailing list archive at Nabble.com. ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Geoserver-users mailing list Please make sure you read the following two resources before posting to this list: - Earning your support instead of buying it, but Ian Turton: http://www.ianturton.com/talks/foss4g.html#/ - The GeoServer user list posting guidelines: http://geoserver.org/comm/userlist-guidelines.html Geoserver-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/geoserver-users