Oskar Linden updated us on this week’s server updates.
The Lindens had planned to roll out the threaded region crossing update that has been in and out of release channels for weeks now. Last minute build errors prevented its promotion. So, Phase I Threaded Region Crossing goes back into QA. We won’t see it reappear for a few weeks.
Coyot Linden says the reasons for the fail are not relevant to SL users. I suppose that means it has more to do with the Lab’s compile and update process than a failure in the region crossing part.
The server maintenance pack rolled from the release channel to the main channel. This is the package with:
- SCR-6 – llGetSubString can produce results that crash Mono scripts.
- SCR-281 – recent release to AGNI (12.02.06.248938 I think) breaks llDeleteSubstring behavior.
- Some server crash fixes are included.
- The ban height for parcels was raised to 5000 meter.
- The fly height change was NOT implemented in the roll.
The ban height they are talking about is explicit bans, those made by estate managers and parcel owners. Griefers were flying above the 800m ban height and dropping griefing devices. You’ll find the visible ban lines do not show up above 800m. That is a viewer thing and some future viewer upgrade will have that change… may be.
Fly height is about 150 to 180 meters. That is going to change. We will see it in next week’s roll to a release channel. The Lindens are carefully testing this change. They expect some flight feathers or other flight devices now used to launch people into orbit after the change.
The change will allow avatars to fly unassisted to 4,096 meters. Any device that adds PUSH to an avatar, to overcome the old limit, will likely launch avatars on an exciting ride. I got to test my flight feather in ADITI. They are no problem. They work fine. The neat part is they are not needed below 4,096 meters.
In the past many configuration changes required the simulator to be restarted. The new servers can be re-configured while running. This should reduce the number of rolling restarts that have to be done. In some cases configuration changes waited for a scheduled restart. So, this change will allow changes whenever needed.
Upgrades will still require a restart. So, Tuesday and Wednesday rolling restarts are still with us.
The process was going to be tested last Wednesday. For whatever reason that didn’t happen. The test is scheduled for next Wednesday. The plan is to roll out the packages to the RC’s. Once those are in place and complete, then they will re-configure the simulators to use the fly height.
This process will also avoid some roll back restarts. The idea is that a feature could be setup to be enabled or disabled by a configuration setting. This would let the feature roll to the main channel. If things went sideways, the feature could be disabled without a roll back.
Next Weeks RC’s
Caleb Linden says the next release channel release will have the following fixes:
- SCR-15 – Add flag parameter [vector slice] to llSetPrimitiveParams PRIM_TYPE flags PRIM_TYPE_BOX, PRIM_TYPE_CYLINDER, and PRIM_TYPE_PRISM
- SVC-680 – Increase HTTP Response body size from 2048 to 4096 or larger.
- SVC-5488 – llGetRegionAgents() — returns list of agent keys in the current region.
- SVC-6894 – Excessive EnableSimulator message spamming to viewers.
- SVC-7307 – creation of scripts gets a nice error.
For a couple of weeks Oskar has been talking about performance improvements from the reorganization of the servers. They call it rebalancing. You may remember they were going to put adjacent regions in physically adjacent servers. We were also going to see a blog post and graphs about the improvement. I am guessing that is not going to happen now. But, we did get some numbers.
Coyot Linden said, “Since the rebalancing ran, we have seen a drop of more than 50% in the number of regions crashing, going from an average of 270 regions crashing per day the month before the [rebalancing] run to 114 since the first day of the run.”
The rebalancing has created much better long term stability. Many of the stability issues that created a need for weekly restarts are gone. The simulator crash rate being cut in half, a 50% reduction, is a huge improvement.
Oskar says he will give us more numbers on the rebalancing change… some day.
Pathfinding User Group (PUG)
The rumor is… well more than a rumor… Falcon Linden said he is planning to start a Pathfinding User Group next week. He things the meeting will be at 4 PM SLT Thursdays in ADITI. Watch for the announcement next week.
I think this is going to be an opportunity for questions and feedback. Once the questions are answered the meeting will be over to the next week… or something like that.
We’ll have to wait and see if we get an in-world group.