2021 SUG meeting week #26 summary

Riding Zany Zen’s Railway, April 2021 – blog post

The following notes were taken from the Tuesday, June 29th, 2021 Simulator User Group (SUG) meeting. The meeting was recorded by Pantera Północy, and is embedded at the end of this summary.

Server Deployments

See the server deployment thread for any most recent updates / changes.

  • There was no deployment to the SLS Main channel on Tuesday, June 29th. At the time of writing, an investigation was underway to determine why some of those simulators on the SLS Main channel due an automatic re-started hadn’t received it.
  • Wednesday, June 30th should see the RC channels updated with simulator release 560819. This includes internal fixes, a fix for BUG-202864, an a further fix for BUG230881 – “llHttpRequest(): HTTP_CUSTOM_HEADER flag is ignored”, which caused an attempt to deploy the internal fixes + BUG-202864 in week #25 to be rolled back.

SL Viewer

There have been no official viewer updates to mark the start of the week, leaving the official pipelines as follows:

  • Release viewer: Project UI RC viewer, version, dated June 14, promoted June 23 – NEW.
  • Release channel cohorts:
  • Project viewers:
    • Legacy Profiles viewer, version, dated October 26.
    • Copy / Paste viewer, version, dated December 9, 2019.
    • Project Muscadine (Animesh follow-on) project viewer, version, dated November 22, 2019.
    • 360 Snapshot project viewer, version, dated July 16, 2019.

In Brief

  • The issue of scripted objects rezzed by other scripted object experiencing a 2-3 second delay in start-up is still being felt, and the Lab continues to investigate the problem. Commenting on the situation, Rider Linden stated:
It is frustrating, and we are looking at a few things that we hope will make the issue better. But no ETA on when to see the changes. I have a test object that has been running for quite some time to try and catch that problem. It’s gone through roughly 3M rezzes at this point and seen one failure (which I was able to correlate with a region crash). 

It has been noted that if the parent object is *not* in the land group, it will work so long as the owner is nearby and wearing the group tag, but will stop working in the owner leaves the region / parcel.

  • Region crossings: LL plan to carry out a “re-evaluation” of region crossing, but this may not happen until 2022.
  • The engineering team was asked what is being worked on (a common question at SUG meetings), and Mazidox Linden (QA lead for the simulator / servers) replied thus:
Plenty is in the works right now.. Most of it is still under wraps, and a fair bit of it is on backend services that you’ll never notice, except for “Oh, I don’t crash as much” or “Oh, my inventory fetches faster/more reliably” (just as random examples of backend work that we don’t generally broadcast).

In addition, Rider Linden re-iterated that a major area of work is upgrading all of the back-end tools the Lab have at their disposal for system management / monitoring (and presumably better leveraging the additional tools / monitoring AWS offers).

  • Some people still seem to be under the impression that texture and object data is passed to the viewer via the simulator. This actually hasn’t been the case since the Lab introduced asset data handling via CDNs in 2014 and later expanded to include all asset types, as Simon Linden noted:
The server doesn’t really do much with texture [or other asset] transfers It sends the IDs and lets the viewer decide what to download or read into the texture pipeline
  • Responding to a question, “Do server still host multiple regions in the same fashion as before uplift? One CPU for full regions, 4 for homesteads?” Simon Linden responded:
I’m going to be really vague to answer that, but it’s both mostly the same thing and also totally new. Comparing a CPU from our old servers to what’s in AWS … from the start, it’s hardware CPUs vs their  [AWS}‘virtual CPU [vCPU].

While Rider Linden added:

That is mostly unchanged… but we’re not looking at the same kind of CPU so it is a bit like comparing apples and oranges.

The use of vCPUs, coupled with Amazon’s overall approach to billing for services, is a big part of why *it’s complex” is a general reply to question on how things operate and are billed within the AWS environment.

  • There have been requests for larger script sizes to be allowed in Second Life. While this hasn’t been entirely ruled out (although there are currently no plans to introduce anything), Mazidox Linden noted:
Larger scripts would probably come with trade offs that would make them unpalatable, like a small limit on the number you could have in a linkset or only region owners being allowed to use them, just as examples off the top of my head.