Rolling restart for 1.24.7 2008 Sep 30-Oct 2

Blog post : http://status.secondlifegrid.net/2008/09/29/post263/

i hope this is policy now because it makes more sense to me to be doing updates early in the week. it gives the rest of the work week to put out fires before the weekend.

prolly just jinxed it. sorry all...

/me thinks the url should be http://status.secondlifegrid.net/2008/09/29/post263/ .

Release notes: http://wiki.secondlife.com/wiki/Release_Notes/Second_Life_Beta_Server/1.24

...............another rolling restart !

if you are going to continue to do this every week please fix the bug that deletes the last 10 or 12 characters in the land description. It is very annoying to have to redo this every week now , contrary to your past comments that you are not doing restarts every week.

Release Notes for Second Life Beta Server 1.24.7 (97922) (September 29th, 200:

Fixes:

* Multiple touch_start events are generated if Left Mouse button is clicked & held
* Three security fixes


What about the bugs that have been on the shelf for 90-120 days ???


Including SVC-2675

SVC-2675 could not be that hard to fix now. Simon, Andrew and Sidewinder are aware of it but still no fix??????

You guys screwed up the economy of SL with the last rolling restart for days and now that it's over you doing it again? Can we please get some breathing time to recover from this fuckup before you try to kill us merchants again?

Seriously, I freaking hate that my SIM's all get rebooted without me having a word in it. In fact it pisses me off that they get rebooted without me allowing it in the first place.

STOP IT!!!

AND PROVIDE SIM OWNERS WITH A DECENT WINDOW OF ACTION LIKE ANY OTHER COMPANY DOES!!!!

Usually my cable company gives me a 2 hr window where they show up. you make it a THREE FREAKING DAY window. This has to stop!!!! You hurting us all.

Dude, they are acting in your best interest here. Did you miss the part of them fixing 3 security issues?

Your a merchant right? uptime > security?
yeah, i think you should be demanding this update sooner and faster, than not at all.

Innocence is bliss.

I don't say this enough.. Thanks for keeping up with the updates.

Can we assume that at least one of the security fixes is related to the recent "steal your session" exploits that have been patched in the latest editions of the client?

I know you're limited on what you can say.. but once they're deployed, I hope that someone would come back and explain to us what these "security fixes" fixed.

Here's the thing, they usually don't want to speak about the security fixes afterwards because they don't want people to know what others were able to do to hack in. After all, what if it turns out that the fix isn't 100% perfect? Also, it could encourage people to look for other ways to accomplish the hack.

I'm guessing that the security fixes are the main reason for this update, since it was expected that once Mono is up and running, we'd be moving to 1.25 and the piles of bugfixes queued up for that. Can't really complain about having LL inserting security fixes as soon as they're ready. The touch_start() issue could potentially break content, or even crash a region in an extreme situation. It's also possible that it was indirectly related to the new llDetectedTouch() functions introduced in 1.24, and thus was part of that project already.

The security fixes are indeed the reason we're doing a roll this week.

The "steal your session" security problem had a patch with *last* week's fix; that's not what's being fixed this week.

Re: the window in which your sim will be restarted once, please read the blog post. It's not three days. It's three different windows on three different days; 1.5 hours one day, four hours each of the other two days. Describing it as a three-day window is misleading, and overstates the situation to appear more dire than it really is.

Why do we do it this way? Second Life is big enough now that it takes a while to restart the whole grid. It has become impractical to do it all in one day, especially since we want to avoid doing restarts within several hours of the concurrency peak (which happens somewhere mid-afternoon SLT). Also, the pilot roll (the reason there is a third day in there) is 10% of the grid, and essential in determining that the live-environment deploy of the new version has not alarmingly increased the crash rate. The plan going into the future will be to leave the first rolling restart in a second-number-increase deploy (i.e. the next one being 1.25) on the pilot group for several days before doing the whole grid. That will give us "bake time" when there are a lot of changes. (When we do rolling restarts with a change in the third number, these are small, scoped changes, and aren't as major.) I certainly hope people aren't going to be posting "ZOMG! The rolling restart window is now a WHOLE WEEK!" when we start doing that....

Prospero I will just because heh
well depends if I'm not feeling lazy that day heh

Then why can't you tell me on which of these 3 days my SIM's are going to be restarted? Because if you don't do that it DOES make it a 3 day window for me.

Can you provide more information on something I noticed in the release notes?

http://wiki.secondlife.com/wiki/Release_Notes/Second_Life_Beta_Server/1.24

For 1.24.6, it states an llEmail Throttle was implemented. However, llEmail has always been throttled, so I'm a tad confused.

Was the Throttle tightened, or was it that Mono had failed to include the Throttle that's always been present in LSL?

Thanks.

After the initial test round and once the process gets going, Prospero will usually say if he's doing odd-numbered or even-numbered hosts firsts. The 2nd day of the RR, he'll do one and the 3rd day he'll do the other. To find out the host number of your sim, TP there then do Help -> About Second life and look for the host name.

Prospero has said in the past, more than a few times, that he doesn't like to give out the odd/even information beforehand because your sim host isn't really reliable. If your sim restarts for ANY reason before the RR, it's got a 50-50 chance of ending up on a different -numbered host.

Personally, just IMO, I don't agree with this reasoning. I'd rather people were told beforehand with a disclaimer that nothing's guarenteed until it actually happens.

I agree with this 100%. Just like Alessandra said, until my sim restarts it is a 3 day window for me too. Knowing ahead of time the even/odd, even if I can't know the 3000 regions in the pilot roll ahead of time would help me a lot. It would then be my responsibility to know whether I am actually an even or odd at that point in time.

Or just make it a standard: After the pilot roll, always make the first half of the grid the odds and the second half the evens. Why couldn't we do that?

Some things I want to make sure people understand:

* You don't have to be there when your sim restarts. Hence, the analogy to the utility service window doesn't really apply.

* Your sim is only down for a few minutes. If it stays down longer than 20 minutes, you should contact support, as something went haywire that wasn't supposed to.

* Your sim will be restarted only once as a result of the rolling restart, unless something bad happens and we have to do a backwards roll-- and we haven't had to do that with the last several rolling restarts.

The only reason I can see the uncertainty about rolling restarts being a big issue is if you have an event planned during one of the four-hour windows. Then, yes, having the restart in the middle of that can be very disruptive. But if it's just you doing things in your sim, or it's just you and a few friends-- then when you get the warning, you need to TP away for a few minutes, and come back in 10 minutes. However, if you're viewing the three days during which time each region will get restarted once as a period when you can't do anything, you're overreacting. Just treat Second Life as normal, and when you get the warning, TP away for a litlte bit.

I think most people do understand that, Prospero.

It doesn't really matter much to me and I don't even bug you (much) with the odd/even question unless I'm really drooling after some SVR fix. That's because I don't use SL for RL work, though.

If I did use SL for work - say I had to set up a meeting with various people spread out across the country/world - I'd be a lot more concerned about it. Getting a bunch of people together at the same time is hard enough; knowing that you might have to send them for a 10 minute break at some random point in the middle is harder; getting them to all come back afterwards and restart the meeting is even harder.

I ain't complaining, not at all, but I do think LL is going to get more pressure to make the maintenance times more manageable (or at least predictable) as you guys get more corporate customers.

Can I get an answer please???

HMMM you said no problems the last few. well i will say your lying there. since the 2nd from last one. there was a major problem on my sim i lost over 1600 us dollars worth of inventory my whole house and stuff in it and 55 zyngo machines. so i would say there was a problem.then to top it off to beable to get a rollback to a time my stuff was there yet i had to get ahold of a sim owner i only see 3 times a month and beg and plead for it cause i being just a property owner in the sim i cant ask for a restart. which would of been alot easyer and cheaper for me.and what makes me mad is i see this right after the rolling start and had to wait 2 days for this to get fixed.even when linden labs admited stuff didnt come back.

Leo Linden Hi Game,

Thanks again for alerting us to this problem with Aloha Cliffs. I checked the logs for the region and they do show that some content did not come back to the region after the recent restart.

The way to get the content restored to the region is for us to perform a rollback back to 09-11 21:47:27 for the region.


We can only, however, perform a rollback if it is requested by the island owner or one of the Estate Managers. If you can arrange for the owner or one of the Estate Managers to submit a ticket requesting a rollback, we should be able to take care of this very quickly for you.

If the owner or a manager requests a rollback, it would be helpful if they could refer to this ticket (4051-5246325) in their ticket.

now thats b.s. you guys did the restart lost my stuff and then say u cant do nothing even when u know stuff is missing from it.

you might want to restate that because you 'did' do a rollback for the last two........

Game -- when that happens, contact your estate manager. If you're on a private estate (as opposed to mainland), they are your first point of contact for questions like that... as Leo told you in the ticket. In general, we don't roll back islands (private estates) without the knowledge of the private estate owner.

Re: the llEmail throttle that somebody asked about a while back-- there has long been a throttle that makes a single script sleep for 20 seconds after sending an email. The new throttle is per user... some were using many, many different scripts to send spam.

Renee -- the issues that get patched in "point" releases (i.e. where the third number increases) are issues that were introduced with the latest major release (i.e. where the second number increases), and security problems. General issues that are fixed or introduced will come with major issues. You can follow their status by tracking the JIRA. If it has not been assigned an internal Linden JIRA, it is unlikely that it is currently being worked on. Looking at the issue you cite in particular, it seems that others have not been able to reproduce it, even though they have been willing to try. I would suggest that you try a less combative attitude, and give people the exact location of a place where you see the problem. Please do not discuss this issue further in this forum, as this is not the right place for it; the JIRA is.

Renee -- you are incorrect. Since the first rollout of 1.24, we have not had to do any major backwards rolls of half the grid (or the whole grid) as we have occasionally had to do in the past.

Did they? I thought the last couple releases required some extra bug fixes but they kept moving forward - I don't think they've had to go backwards for a while now..

ppsst Prospero ya might want to change the time to PST on the grid status page

(stupid round earth)