Posted By: Vincent Sylvester
Posted On: 2021-03-02 03:17 AM
Some of you may have already noticed this, lately from time to time teleports will fail effectively getting stuck on whatever region you are on. While we have already taken measures to combat the issue, we would still like to know whenever and wherever issues regarding teleports occur. From the initial investigation the cause is a change in core OpenSim, which, curiously without clear indication as to why, causes the presence transfer from one region to another to fail. Resulting in effectively jailing a user on a region with these changes. Largely said changes were rolled back for builds after 717 and are not present in builds before 690. As these are changes made by core we have been in contact to establish the reason for the failures, being it seems clear the changes are the cause. If you find yourself on a region that does not allow you to teleport anywhere else, please send staff a message or let us know via a ticket of the region and time of failure so we can investigate and hopefully gather enough data to figure out a fix.
In future we plan to manually implement the core changes that are seemingly innocent and avoid the parts that could be the cause, though ideally in order to maintain compatibility and actually take advantage of the supposed fix we would like to use the entire set of changes. However for this they would need to be either fixed or confirmed working, pointing the cause of the issue elsehwere which currently does not appear to be the case. For those interested in the technical side of things you can find the related commits that seem to be cause of the issue on the OpenSim Viewgit
commits from 2021-02-12 to 2021-02-13.
We hope that in time, through confirmation of the issue, it can be resolved and that the actual cause is not due to changes made on top of core OpenSim, which became necessary some months ago due to the addition of caches for presence and user information, which unfortunately clash with ZetaWorlds load balancing setup as this would mean either full reversal of those, currently there is only a minor patch to allow disabling of caching, or switching to a singular grid instance for presence and user data, which reduces load capacity and creates a higher risk of failure should a single instance crash.