Server

We got several complaints this week. I’m not sure how many of them are real. But, when the foundation shifts everything shakes. The HTTP Inventory Services made it to the main grid. That type of fundamental change usually comes with some problems.

Critical OS Updates Complete

Oskar Linden announced that about 3 PM Thursday this update process completed. All LL servers have now been updated.

The additional restarts for the COS updates are over.

It appears this update may have partially or completely solved SVC-5040 – Script performance dropped on regions deployed on class 7 hardware since rolling restart.

Read more

This meeting of the Beta Server Group was Oskar Linden’s 100^th office hour/user group meeting. Lots of fireworks. Oskar started dealing with server QA two years ago. At that time there was a server roll out every quarter. Oskar feels most things have changed for the better. Post a congrat in his FEED.

Main Channel

Tuesday the Fast Scripts update rolled to the main grid. The upgrade improves the performance of a number of script functions. The functions improved are those that do NOT change the properties of objects, so mostly llGet…() functions.

Read more

I missed the Simulator User Group meeting. But, I just saw the meeting minutes got posted. There are a few interesting things that came up in the meeting.

TimeWarp

This is the Linden name for the problem of simulators stalling for 30 to 60 seconds, which knocks people in the simulators’ regions offline. Since the operating system kernel upgrades on the ADITI grid that problem has disappeared.

This week, Oct 3 to 7, The Lab will be upgrading the kernel on all the servers running simulators. If I understood correctly, that upgrade will also be applied to many of the support services servers. This means the week will see lots of region restarts. See: Critical OS Software Upgrade — Region Restarts

Regions will see at least 2 restarts and if unlucky 3.

Read more

Tuesday a new version of the server rolled to the main grid. Only one high priority fix made it to the main channel this week. This version contains the fix for SVC-7283 – llRemoteLoadScriptPin and llGiveInventory fail when installing non-full perm items into worn attachments.

This problem was causing the permanent loss of scripts in worn objects. As it was breaking existing content, this was a top priority to get fixed. It was considered enough of a priority that it was placed on all three Release Channels to assure we would have a fix this week, provided the fix passed testing, it did. See Second Life Server News Week 38 for more information.

Some are upset that it took 2 weeks to roll out this fix. The plan is to test and test and be sure a fix actually fixes the problem without breaking something else. That means a week of testing in ADITI and a week of testing in a release channel. The Lindens thought this fix important enough to place it on all 3 RC’s. If it had failed it would have failed all the release channels and delayed all other fixes in progress. Their concern and urgency on this issue pushed them to use all 3 channels. But, for some that is not good enough. That’s life.

Read more

There was lots of interesting news at this week’s meeting. Check these things out.

Roll Outs

Surely by now you know the main grid gets new server software Tuesday morning. The Release Channels get new software on Wednesday. This week the main grid got the the Server-Maintenance update running on Blue Steel.

Data Collection

The Lab is doing lots of analysis of the regions to find bottlenecks and problems. This roll includes new data collection features. One of the more confusing and hard to find problems is labeled TimeWarp. The Lindens think they have found the problem, at least where the problem resides. The additional data collection will provide more information on many aspects of the Second Life servers and reveal more about the TimeWarp problem (more on that later).

Unexpected Prim Return

The addition of Land Impact costs has created some surprising situations where region prim limits are hit and prims get returned. Changes made by scripts can change object shape and trigger a Land Impact value and trigger returns.  A feature has been added to the Linden Scripting Language (LSL) to anticipate such changes and prevent them. I think the idea is to return an error message in the script error channel rather than allow the change and trigger returns. Nice idea.

Read more