We’d planned on doing server software updates last week, but due to the discovery of a bug this slipped into this week. We took that opportunity to pull in some fixes pertaining to inventory loss, some of which take effect immediately and some of which (which address only perceived loss, not actual loss) will become active when future viewers are used.
We’re going to play the deploy “by ear”, meaning roll it out to a small number (hundreds) or simulators first before deploying to the whole grid. Here’s a rough schedule:
- Thursday AM: Rolling restart of ~500 regions [Done]
- Thursday PM: Rolling restart of ~1500 regions [Done]
- Monday AM: Rolling restart to the whole grid
We plan on updating the regions running the Havok4 Server Beta code in parallel with this early today. Sidewinder Linden will post with an update soon.
As usual, the rolling restarts will occur in “random” order across the grid (they go by hardware arrangement rather than region arrangement). A 5 minute warning is given before each region is restarted.
[Update 2008-02-28 11:06] The deploy of 18.104.22.168913 to 510 regions is complete.
[Update 2008-02-28 17:54] The deploy to 1537 regions is now in progress
[Update 2008-02-28 17:30] The deploy to 1537 regions is now done.
[Update 2008-03-02 20:06] The deploy to the whole grid (exempting those regions running Havok4) will begin tomorrow morning around 5AM. Note that even the ~2000 regions that previously received a deploy will receive this deploy; we have found, fixed, and tested a small problem with those deploys. While this is inconvenient for the ~2000 regions to be restarted, our gradual rollout saved the other ~15,000 regions a double restart as a result of this discovery.