Please refer to the server deployment thread for news and updates:
- On Tuesday, 26th November, servers on the Main SLS channel were updated with server maintenance release 2019-11-15T21:13:13.532828, previously deployed to a release candidate channel. It comprises:
- Improves crash detection during shut-down.
- Fixes some race conditions with LSL scripts that could cause them to handle the same event more than once during a roll.
- Includes updates to improve simulator security.
- On Wednesday, 27th November, there should be a single RC deployment 2019-11-19T22:26:38.532992, which is functionally the same build as 532828, but compiled using updated server build tools.
- Given the issues with Tuesday deployment (see below) the status of the deployment was in a state of flux at the time of writing.
Tuesday Deployment Issues
The Tuesday server updates required an extended deployment window, together with more restarts of the updated simulators than is usually the case. Given the simulator version had been deployed to RC servers in week #47, the issues encountered with the deployment took LL by surprise:
This is one of those situations where we are left trying to figure out what went so wrong; we don’t know of changes that would have caused this much churn. That is the big difference between our Tuesday and Wednesday updates. Tuesday is more than 2x bigger, and divided up differently. Seeing something happen and knowing why it happened are two completely different things. Given that last week was OK, I suspect it is a scaling problem however.
The real frustration is that there isn’t a good way to test or simulate the scale where the issues happen. Updating one or two servers can be perfectly fine; but then do a few hundred and something goes up in smoke.
– Simon Linden, commenting on the November 26th deployment issues
Currently, the Lab is digging into what may have gone wrong (e.g. by checking server logs, etc.). They have already tracked down what might be a contributing factor, but the overall root cause analysis will take time. However, as the issues appears to be with the deployment process itself rather than the updated simulator code, at this point in time it is unlikely the simulator update will be rolled back.
The following viewers were updated during the latter half of week #47:
- November 22nd:
- November 20th
The remainder of the viewers in the current pipelines are as follow:
- Current Release version 18.104.22.1682299, formerly the Ordered Shutdown RC viewer, dated November 4th, – No Change.
- Release channel cohorts:
- Project viewers:
- There are concerns over how long names might be cached within services, how changes might be tracked by external databases (e.g. customer lists for things like redelivery terminals), etc.
- To help ensure user-developed tools, etc., can keep up with Name Changes, Linden Lab plan to offer a REST API (REpresentational State Transfer) that can be used to perform a name<-> avatar key translations and that will always return the same key for any name an account has ever had, and the key will always return the current name.
- Creators are again reminded that their products and tools should utilised the avatar agent key, no avatar names.
- LSL enquiries on agent key will return the current name for an avatar, not any past name(s).
- There was concern that Name Changes could lead to issues in tracking griefers. However, as was pointed out in the meeting:
- There are fees associated with Names Changes – signing-up to Premium and then for each change. These are unlikely to find appeal with griefers.
- Griefers already create multiple accounts, and this will remain likely remain their preferred means of causing upset.
- BUG-216397 “llName2Key, llRequestUserKey both accept a lone “R” as a surname for “Resident” agents” has been raised with the Lab again, and has been marked for inclusion in the next internal maintenance fix for Name Changes.
- Once again, a “fee” figure was mentioned at the Simulator User Group meeting (US $20) – but it is important to note this is only a guess on the part of a user, and not any form of statement from Linden Lab. So If you hear it anywhere in the next month or so, ignore it.