Please refer to the server deployment thread for news and updates:
- The SLS (main) channel on Tuesday, November 5th, leaving it on server release 2019-10-26T00:06:48.532192, previously released on October 30th and comprising:
- A previously released hotfix to fix teleports being 5%-7% less reliable.
- A change that makes the simulator take a little bit longer to report as “Up” to the Lab’s internal tools to more accurately reflect when residents can actually access a region.
- There is a single RC deployment planned for Wednesday, November 6th: version 2019-11-01T18:02:37.532376, which is essentially a rebuild of 532143, originally deployed on Wednesday, October 30th and containing the above updates as well. No release had been released at the time of writing.
Note: a second RC deployment, 532377, comprising an internal update and a script to monitor additional region data during rolls was cancelled after it was discovered in QA it broke script uploading.
Linden Lab request that if anyone sees messages about “Unable to upload 12345678-90ab-cdef-1234567890ab due to the following reason” on a version that isn’t 532377, could they please file a bug report and provide full information of when, where and how they encountered the message.
The Ordered Shutdown RC viewer was updated to version 18.104.22.1682299 on Monday, November 4th. At the time of writing, the rest of the viewer pipelines remain as follows:
- Current Release version 22.214.171.1241811, formerly the Voice RC viewer, dated October 18, promoted October 31 – NEW.
- Release channel cohorts (please see my notes on manually installing RC viewer versions if you wish to install any release candidate(s) yourself):
- Project viewers:
- Project Muscadine (Animesh follow-on) project viewer, version 126.96.36.1991949, October 28.
- Copy / Paste viewer, version 188.8.131.521844, October 21.
- Legacy Profiles viewer, version 184.108.40.2060836, September 17. Covers the re-integration of Viewer Profiles.
- 360 Snapshot project viewer, version 220.127.116.119111, July 16.
There are continued reports of simulator performance issues – notably with Mainland regions. The most common description of the problem is that over time, script performance on Full region simulators declines to the point where only around 20% or so of scripts are running per cycle, and the simulator has no spare time, forcing the region holder to submit a ticket requesting a move to a different server. This tends to resolve the problem – at least until the next rolling restart, when the performance degradation starts again.
There appears to be no discernible reason for this. Rider Linden has been making improvements on how scripts – particularly idle scripts – are handled, and at the meeting, Simon Linden noted:
We’ve been looking at scripts and performance and there are some changes in the works that will hopefully help … those are on about 15% of the regions now, and will expand to about 20% tomorrow.
…Rider’s work looked at a bunch of the event distribution code like that, and he made it faster with re-writing it and getting rid of brute-force searching and so on. Of course, that changed some script timing and killed some kittens and pot plants. Sometimes it really hits me what a strange job this is.
The issues have led to speculation that the Lab might be stacking more regions on their servers or are reducing the server resources available to each simulator. Responding to an enquiry on this, Simon indicated that the Lab have not changed either for years. However, he did note the fact that the servers the Lab uses do change, so something may have been introduced to impact performance.