Etchification Rolling Retart 2009/02/17-2009/02/20

I worked for a firewall company and could run SL at work just fine. Unless your corporation is blocking streaming for bandwidth reasons, or has ALL ports which are not authorized locked down, SL should work just fine.

Also, one of the better usage models for meetings in SL is bringing together distributed teams who are working from home.
.

There are already businesses doing simple conference facility rentals. A model better suited to success is an "event management" business, which gives you avatars and a certain amount of other handholding for users who are not yet in SL.

Statistically speaking, is it not far more likely that a meeting that takes 120 minutes would go down during the meeting than during the 15 minutes before it, when attendees are arriving and you could feasibly move to another sim? Having the sim go down in the middle of a meeting where there are several new-ish users would not be spiffy.
.

I can believe that. A lot of security companies have a fairly laid-back attitude to security.
Default deny is a moderately secure model. When I set up our firewall in the mid '90s I went proxy-only as far as I could, so not only was it default-deny, it wasn't even transparent. And our DMZ was default-deny and proxied as well, with each internal customer in its own DMZ. Which turned out to be rather useful when Code Red hit a Windows-based database server and couldn't make any outgoing connections.
If the protocol was firewall-friendly that could be handled via a virtual proxy.

...but rolling restarts *are* announced ahead of time. The only exception would be an emergency (e.g. a security patch), but note that RL also sometimes has to deal with emergencies (e.g. power outings).

If you are an estate owner and you've scheduled an event that can't be moved, you can contact concierge and ask to be exempted from a particular day of the rolling restart. Only do this for events that really can't be moved; if we get more than a few people doing that on a given rolling restart, we won't be able to keep doing it. Yes, we do hope in the future to have a way where you can *always* specify times to be exempted from rolling restarts, so that you can have more predictability. We have talked about it, but at the moment we don't have such a system.

You do not describe something I recognize.

A typical rolling restart has a pilot roll announced with a 2-hour window, and then a full rolling restart up to a week later with two 4-hour windows, during which each region will get restarted. Not three days; not even two days. Two 4-hour windows, one on one day, one on the next day.

I keep saying this, but people keep insisting that they want to view Second Life as being completely unreliable over the entire time from the beginning of the first window to the end of the last window. Why?

I think Taff is referring to event scheduling, Prospero.

If you were in, say, the wedding business, you really don't want to hear that your region will be down "sometime" during a 2-3 day period. It's hard enough to get a big group together at a certain place at a certain time - having a restart roll over them and trying to get everybody back and restarted is even harder. Same goes for any kind of event management, I guess. That and things like business meetings.

Would LL consider changing the way sims restart so that during the RR period, and maybe a day before, sim were pinned to the same host? I understand that'd mean that a crash would take longer to recover from (how much longer I have no idea) but would give people a far better idea of when their business might be impacted.. Just a thought..

Thanks for the answer, Prospero. It's great to be able to discuss this with you.That would be SO helpful. I'm sure LL's new hires responsible for building your enterprise sales will be happy to hear that too The full restart in this thread was scheduled with 4 days notice, 2 of which were over a weekend. The previous one: http://status.secondlifegrid.net/2009/02/05/post473/ was 3.5 days notice, again over a weekend. It is difficult, with 2 business days notice, to reschedule RL meetings that happen to fall during those windows. After all, the competing platform for RL business meetings is not Oklahoma during tornado season; it is Webex, which has never requested that I reschedule a conference.Hmm yes I see that this restart, being 2 12-hour windows plus another long one, is atypical. My apologies. But I don't view SL as unreliable during this entire period; I view any single sim as vulnerable to restart during any of the non-pilot windows.
.

But it's *not* "sometime during a 2-3 day period." It's "sometime between 7 and 11am Tuesday or Wednesday." A 2 day period is 48 hours long. That's very different from two four hour blocks.



That would create a host of secondary problems... on the whole, it would lead to less grid stability. As we are right now, we're very resilient if a host dies. Yes, some regions have to restart because their host died, but they can start on any other host. At the very least, if they were locked to hosts, then they would require manual intervention, and would not come up as fast.

I'm Innovation Manager for a well established ITSM consulting and training organisation. We are taking the first steps to using SL as a vehicle for extending our business reach. The months that I have been in SL have shown me a textbook example of a company attempting to deliver an IT service from very low on the ITSM maturity scale.

This isn't too surprising, most IT organisations start out like that. For a while it is innovation, innovation, growth, innovation and all the while paddle like hell to try to keep it afloat. It takes time and experience for organisations to move beyond that and to realise that stability, service and delivering business value to the customer are the real game, without losing the ability to innovate. And the term "maturity model" gives us at least a hint that organisations, like people, start out with a low level of maturity. Some IT organisations (like some people) don't make it to maturity, they die first.

In any event it looks likely that it will be a while before LL gets the level of maturity necessary to see eye to eye with the likes of Atashi Toshihiko who see predictability and business value as the drivers. A hard won lesson from helping organisations improve ITSM maturity is that it is VERY HARD to make big leaps in maturity in a short time. From where LL is today (let's call it CMM level 1) to get to where Atashi is asking (CMM level 3) won't happen over night. I just hope that LL survives long enough to gain that maturity.

Lindens, feel free to give us a call, we can help!

I think its very unfair of SL to be doing these rolling restarts during the evening hours they should be doing them overnight and late night hours not when alot of events are happening

One avatar's evening is another's morning. Welcome to the world!

Patrick: your article is still incomprehensible to people who don't know the acronym "ITSM".

In psychology its called the primary/recency effect: People tend to remember things that they percieve at beginning and at end of a communication better than whats in the middle.
So if you say: "there is a 3 day outage, every region will be down 5 to 10 minutes, it will be Monday to Wednesday" things that are rememberd best are : "3 day outage-Monday to Wednesday".
If had to do scientific research about that I would test for the hypothesis that people percieve the outage shorter when putting the sentence in a different order, like:
"your region will be down some minutes while the outage frome Monday to Wednesday, it should be up again after about 5 to 10 minutes".

True. Try Wikipedia if you are interested.

I'm not interested in what ITSM means, actually. I'm just pointing out that after you edited your comment to make it more obvious, you still didn't make it obvious. I thought that was amusing.

I didn't say obvious, I said ambiguous. I didn't mean to be obvious.

Ah, well, you certainly succeeded in your goals then.

I cant log in..it all comes up and then when I log in it says sory unable to connect to secondlife,despite our best efforts,somthing unexpected has gone wrong.Then gives me the status site for sl....well thank-you for ur time.

This thread is about the Feb. 17-20 Rolling Restart.

If you want to help, use JIRA and official mailing lists. I can provide you with more information about these tools if you need it.

If you want to offer partnership and commercial services to Linden Lab, please don't use this forum - you're really off-topic in the Server Deploys forum category.

I'm not sure if an ongoing Rolling restart could cause that error.
Did you try logging in a different region? You can enable the "show start location" option in Preferences, then insert a region name (or SLURL) at the login page. If you still get that error, try with another region name; if you get it again and again, I'd say the reason is NOT the ongoing Rolling Restart.

I'm not familiar with the acronym, but I did clearly understand everything else that PatrickJ said.

Arguing semantics and ignoring direct questions appear to be the LL party line. I also understand precisely now what Prospero is saying. SecondLife is a game and not a business platform.

You might want to make sure that M, the other top guys, and the PR folks are brought up to speed. They keep making these comments on the blog and elsewhere, alluding to SL being about business and education.


Personally, this quote sums up very nicely what I have failed to convey myself:Of course all of SL is not unreliable, but any one particular sim *is* for those hours when the restart is scheduled.

-Atashi

I can't enter to Sl when i try to log me appear a small windows saying "Login failed Unable to connect to a simulator" why the SL show me that?

Well I think restarts are needed and you will always hit someone in his event. And yes, they always come when I am online in my evening time. Only thing I don't like with it, is how SL reacts then, is just even slower than with a high AV load. But I won't complain. I only hope you don't catch more bugs with the different compiler versions or mess it up again

Prospero have they started yet?

there's a small change of plans... watch the updated status page