Rolling Restart to deploy 1.27.0

I get it that the weird failure to populate object contents was a disk-full condition on the cache, but not sure if that applies to this:The thing is that I'm seeing that behavior, too (and not caused by a full disk in my case). I don't think I see it being a lot worse than before, but the Release Notes mention: which might lead one to expect, well, "improvements"--and I'm sure not seeing them, but rather just a steady gradual degradation for months.

Is anyone plugged-in enough to plans for the texture loading pipeline to comment on whether any serious relief is on the way?

It's gotten so bad that I've changed my building and buying approaches now to completely avoid sculpties and to constrain myself to the tiniest palette of textures I can possibly use. That may be instructive experience, but even taking such steps to an extreme, anything involving texture downloads is painfully slow.

Well you were right, something was filling up. Glad you figured it out.

What's wrong with that now?

Back story: Every day, I do rounds of all our satellite stores to do a visal check that everything is in working order. Typically, at least one time during rounds, a TP will fail. And also typically, I try again right away.

Till today, that was fine - sometimes it would work again right away, other times not (eg, destination region is down, or full or whatever).

Today it happens again: TP to my next destination failed. I try again right away. But THIS time I get a message about TP being unavailable - and am LOGGED OUT!

I read the release notes about throttling being to prevent griefing and/or to help grid stability (which is it, btw?), but logging us out is extreme. You're interrupting my work that way! You want to inhibit TP for a brief time? Fine, do that, but why log us out? That's going too far.

it appears to be related to some viewers not handeling the throttled TP condition right, not sure what viewers suffer from this, I have not noticed it myself, not even during RFL with some sims suffering badly from handover delays (the sim RFL Strength even blinked in and out of view at the end of the relay)

I'm sorry I wasn't able to respond when you all were wanting information. I was actually on vacation (to attend a friend's wedding, scheduled far in advance), so others did the full deploy to agni. I will try to work through all the posts and respond to those where it would be constructive to do so at this point, post-deploy.

TY, Dante - grats to your friend and hope you had fun at the nuptials!!

Getting some comment from LL about the recent plague of sim stalls when people TP in would be great, either here or a blog or somewhere. I know people have demonstrated this to concierge Lindens over a month ago. Some more info from LL would really be nice....

We might be able to come up with an app that does something like that, but I'm not sure it's a higher priority than a lot of the other work we have to do. I don't know if people get tired of hearing this, but you could always make a jira requesting that we provide an interface to the information you want and try to get people to vote for it.



There is a grid-wide setting that lets us lock regions to the simhosts they are on, but it's not good to use for long periods of time because if a simhost crashes, the regions that were on it won't start on any other simhost.

Personally, I think I would try to avoid scheduling events during any of the RR N-hour windows and in the early morning and evening (Pacific time) on the day or two following the scheduled deploy days, in case the dates need to slide a little for some reason, as they did with 1.27. I realize everyone's circumstances are different, though, so that might nt always be possible.

Ok... I'll see if I can come up with a JIRA that makes sense.

With the way LL appears to be targeting corporate cash and in-world business meetings and such, I'm surprised this isn't something you guys are doing anyway. Just being able to pin down which day a region will get rolled on will cut your planned downtime stats in half. Seems like low-hanging marketing fruit to me tho I'm sorta back here in the cheap seats .


Yep. I was suggesting a tweak to that.. Instead of having a binary switch that would tell regions to stick to the same host or not, add a third value that says they can move to different hosts but that if they are on even-numbered host, they should try to come back up on an even-numbered host. This gives people a much better idea of when the roll will get to their region without adding as much risk as being pinned to the exact same host.

/me is tempted to state that this would increase predictability but I think I'd get stuff thrown at me from Lindens and residents alike...

Are you still seeing this behavior?

Is this still an issue? If so, is there a jira number you can refer me to?

Dunno. I didn't personally see this problem - just saw what seemed to be a lot of resident comments about it all of a sudden.

The sim stall problem, unrelated to textures, was still very much happening last night.

/me adds http://jira.secondlife.com/browse/WEB-1224 for the rolling restart stuff I babble about in posts above.

The 1.27 upload problem is https://jira.secondlife.com/browse/VWR-14724 (yeah, misfiled as VWR, but that's the most active of them).

We're investigating an issue relating to region neighbors flickering and border crossing failures that's separate from the known issue in 1.26 with region neighbors flickering. I'm not sure if there's a pjira for it, but it's being looked at.

I'm not sure what you're reporting, since the link limit for two default prim cubes is 4.598 meters, per this wiki page:

http://wiki.secondlife.com/wiki/Linkability_Rule

Can you clarify what you're suggesting has changed in this release?

Dante Linden, i seriously suggest you and your other linden lab cohorts get out of you little boxes and into the rest of secondlife. Ever since this last supposedly "successful" rolling restart, there have been nothing but problems. Every few seconds we are hit with "fail to upload" messages, there is serious lag time for everything more times than not, the network and server are constantly flickering between being "normal" and poor connection. Even when things are "normal", there is still serious lag time. This entire thread has listed nothing but issues with the latest deploy and rolling restart and we are yet to hear or see anything being done about. It is no use playing ostrich and dumping you head in the sand. there are serious issues with Secondlife that have not been addressed. And everytime you fix one thing, you break a whole lot more other things. This has been a constant. get in people who actually know what they are doing - because the tech people you have are performing like amateurs. honestly. No one wants a faulty product yet that is what you and the linden lab teams have given us - a faulty product.

now what is the bet that no linden will address this particular message?

I shouldn't have speculated, as I know nothing about the inner-workings of third-party viewers.



I can't reproduce this. I've tried before and failed to repro it, and just now, I crossed the threshold and tried teleporting 15-20 times, and each time, I got the threshold message and remained logged in. What version of the viewer are you using?

If you look at previous release notes, there was a new section for every point release. We didn't make a new point release for the fixes we added, perhaps since the first full deploy hadn't completed before we implemented additional fixes (I wasn't here for those discussions, so that's just a guess). I'd expect that we'd make new sections for each point release, as we've done in the past.

TY, Dante.

Search should be fixed (SVC-4563), and I just asked our support group if they're receiving any reports of map or traffic problems, and they are not. Are you still seeing problems with map and traffic updates?

Goes along with I tried to post a event today on my adult rated sim and couldn't because the events listing saw it as a pg rated sim so now it's back to mature until LL gets things straight

Why does a failed tp kick in the threshold?
as in: when i get a failed tp twice in a row i get the threshold message and can't try to tp for a few minutes?
Why would a failed tp even be considered in the count for the tp threshold?

oh and the link on linkability rules you gave is to an empty wiki page, but it was never this close, it was always 30m between prim center to root regardless of the prims sizes. at least I never ever had a problem with linking prims until 1.27. May just be me but this is something i do quite regularly and its something I noticed immediately.

One side question: Is Zindra "green" because the sims are actually running on the class 6 servers announced in this blog: https://blogs.secondlife.com/community/technology/blog/2009/05/14/the-greening-of-second-life
Is this part of the reason so many Zindrians are complaining about lag?

There are two issues here.

1) The logout issue is a bug, but it's not one I've been able to reproduce. As I mentioned earlier in this thread, I always get the threshold message and remain in-world, and I've tried teleporting about 20 times after reaching the threshold. If anyone has a reliable repro, let me know.

2) Your failed teleport should not count against the threshold. If you want to file a pjira about it and point me to it, I'll import it and get it assigned to someone. If you don't want to file one, I can just file an internal one to get it resolved. Let me know either way.

The threshold is meant to reduce teleport abuse, which I guess counts as griefing.

The wiki page is: http://wiki.secondlife.com/wiki/Linkability_Rules. It's really quite good and has been there for some time (almost 2 years). There have been number of discussions here in the forums about it.

AFAIK, the link distance has never been "anything and everything within 30m" but somebody like Andrew Linden would really be the one to ask.



When we asked about the performance of zindra regions - concern over LL stuffing all the high-traffic places on the mainland into a relatively puny number of regions - we were told that they would all be on class 5 hosts.

A truly horrible ansswer to the question that was being asked at the time but actually does answer your qusetion.

I agree that it shouldn't, and I just made a reply about that issue.



If Meade's link also takes you to an empty page, try going to the main wiki page and searching for "linkability rules" to see what I was referring to.