SL Server Update 10.10.5.x

 

Server room

Server Room - Image by Torkild Retved @ Flickr

 

Blue Steel and Le Tiger Release Channels (RC) were both running the same RC last week. They had problems and are going into another round of testing with new fixes applied. However, Le Tiger is getting a different RC for testing. This is the first time the channels have had different candidates. This is sort of building the pipeline and then filling it. 

In the test channel areas purchases and land purchasing were failing. Worse the servers were losing avatar session ID’s when an avatar tp’d out. I suspect this forced a log off or an avatar stuck in teleport.

Blue Steel RC was updated to server 10.10.05.211372. The release notes are here.

The last server rollout 10.09.21.210146 is doing well. That has different meanings to different people. The known bugs are still known bugs. The fixes added are holding and working as expected. But, the hings they hoped would be fixed in the current RC didn’t make it to the grid. So, we have to deal with the same problems this week as last week.

Oskar found a problem with MOAP and is recommending adding beacons for MOAP objects. They have the code for that. It just has not been merged into the current builds.

Feedback is SIM crossing are better. Note card and script editing are worse.

September 30 was Babbage’s last day. His work on the memory allocation problems made it onto the main grid on his last day. He is reported to have been excited to see it make it to the grid.

Stats are not in but empirically there is a high level of non-deliveries from the SL Market Place (SLM). To get an actual fix rolling, one needs to file a JIRA with lots of info when a delivery fails. Right now it is vague reports. In a good report LL is given information about; which av, what time, from which region, which item, which seller, and machine info (Help About SL).

Because of an LL mod of apache2 (a component used by the server software) failing and not triggering a restart the 10-5 rollout was not expected to make it and didn’t.

Meshes

The Beta team is estimating the number of people that will engage in the Open Beta of Meshes. This should flood ADITI. If you are planning to participate, make sure you can get in now. If you can’t it takes 24 hours for the system to process the changes needed to make it in. Support is not likely to be much help if you have a problem. Do it now before the servers over load.

If you have not been into ADITI in some time, the preview grid likely won’t let you in. The wiki says you will need to change your password and wait 24 hours to get in. It has something to do with how ADITI pulls data from the main login process. That is supposed to be fixed so it is easier to get in.

The regions in ADITI that will support mesh are regions names starting with ‘Mesh’, per Maestro Linden. You can find the region names here: MeshRegions

Region Conductor

This is the part of the system that restarts down SIM’s and assigns regions to various servers. It is being improved and should result in less overall lag through out SL.

Region Console

If I understand correctly it will appear with Mesh. Maestro Linden says, the console will have ‘Restart a region’ (for estate owners/managers) and ‘toggle terrain physics mode’ (a new ‘tri-sampled’ mode which offers higher physics detail than usual).

Scripting

We may see some great improvements to scripting performance as KCP (Kelly’s Crazy Plan) moves forward. Ambroff appears to have found the Mono script freeze-on-rez thing. This will mean Mono can be used in avatar attachments without serious SIM crossing problems. Also rezzing Mono scripts will no longer freeze the SIM. We may see these changes in the second or third rollout from now.

[15:39] Gooden Uggla: mono scripts have been allocating 64k no matter the script size
[15:39] Kelly Linden: there are some specific bugs around adding a mono script to the world, from rez, tp or region cross, that we are addresing.
[15:39] Latif Khalifa: Gooden, not true
[15:40] Gooden Uggla: really? when did a dynamic foo9tprint start?
[15:40] Latif Khalifa: Gooden, since mono was released
[15:40] Kelly Linden: when mono started, gooden.
[15:40] Gooden Uggla: awesome
[15:40] JB Hancroft: lots of urban legend/rumors around mono
[15:40] Kelly Linden: Average mono script in SL uses ~9-10k
[15:40] Gooden Uggla: and now it’ll work
[15:40] Latif Khalifa: yeah, because reporting tools report “reserved” memory not actually used mem
[15:41] Joel Savard: kelly, even though the real server ram allocation is dynamic
[15:41] Latif Khalifa: so people think mono scripts uses 64k
[15:41] Joel Savard: ^^ latif – but it “looks” like they use lots of ram?
[15:41] Joel Savard: in the inworld tools?
[15:41] Latif Khalifa: Joes, yes, looks to the tools
[15:41] Gooden Uggla: so is that memory actually usable?
[15:41] JB Hancroft: so reserved doesn’t actually contribute to VM / swap memory pressure?
[15:41] Latif Khalifa: JB, nope
[15:41] Latif Khalifa: not until it is actually used
[15:41] Kelly Linden: the script-limits reporting tools report the potential max memory the script could use.
[15:42] Gooden Uggla: can that get fixed too please?
[15:42] Gooden Uggla: *smiles*
[15:42] JB Hancroft: but mono does share code space, correct? so if you have many many of the same object rezzed in a region?
[15:42] Kelly Linden: for some definition of “fixed” eventually.
[15:42] Kelly Linden: jb: yes.
[15:42] Joel Savard: *chuckles*
[15:42] Latif Khalifa: it would be very useful to have tools report actual memory usage in addition to reserved
[15:42] JB Hancroft: thanks Kelly
[15:42] Kelly Linden: jb: but that was also the source of the biggest lag bug with rezing mono scripts

 

Leave a Reply

Your email address will not be published. Required fields are marked *