Poll: should linden labs close down the grid for a longer maintenance, a week or two?

Should LL close the grid completely for a week or even two, if that gives them time to _really_ fix thing, instead of plugging holes in a running grid with duct tape and balin wire?


and before you say "nooo that would hurt my business!!!111", think about how much the current state hurt your business every day, for an unknown period of time ahead.

Only if I wasn't charged tier during that time.

Agreed with Plato.

They wouldn't be able to fix things in a week or two.

I've been one of those people who has shrugged off weird SL thing after weird SL thing...believing they would get it together fairly soon but I'm getting frightened and reassessing my involvement.

SL feels like one long stumble without recovery.

A longer downtime would help nothing, as most of the problems occur under high load on the grid, and if the grid is down they cant find out about it - they actually need us to see what is going wrong and then try to fix it.

Only if there was a reasonable expectation that the problems could be fixed within that timeframe.

But I don't think there's anything indicating that this is the case.

The current problems with the inventory database and assets server probably needs a complete redesign and rewrite of the code to be fixed.
It's clearly not up to "industrial strength" at present time.

Running a huge database and still maintaining absolute integrety of it's contents is indeed possible, but for some reason it's not happening here in SL.

Indeed, they try to pull a Mac truck with a middle-class car. Its possible as long as there is not much load on it, but the current set up is meant to deal with 100,000 concurrent users and has massive problems with half of it already.

Stock up on batteries!!

You WILL NEED more batteries.
Various sizes.

Most of those toys that you'll be playing with instead of being in SL are "batteries not included".
Get batteries.

Agreed.

That's why they don't do it. Because they can't charge for a service they aren't providing. I don't know the letter of the law but I also believe the Linden TOS states that they can't charge tier for days when there is a certain amount of outage.(I think if it is over 24 hours)

Regardless of the TOS there are trade laws regarding what a service provider can charge on days when they aren't providing the service.

better ask

would the server been up if no lindens were working today?


they fucked up the code *again*

If they were to close to fix everything wrong with sl, it would remain close for about a year's time.

NO!
this should all be done on the fly.....
maybe a few hours here and there, it happens.....

the goal should be 5 nines, up time of 99.999%
impossible perhaps.

--------------- Offline per day --- per month --- per year
99.999% --- 00:00:00.4 --- 00:00:26 --- 00:05:15
99.99% --- 00:00:08 --- 00:04:22 --- 00:52:35
99.9% ---- 00:01:26 --- 00:43:49 --- 08:45:56
99% ---- 00:14:23 --- 07:18:17 --- 87:39:29

but it seem they are close to the 99% mark, at least during my time online, so i cant complain

At the risk of sounding off all of the bitheads out there, I have heard that SL uses MySQL as a database engine. Reality check. AHEM. Problems have gotten worse as load increased.

Time to consider more power under the hood people!

They can providing they only charge me the relevant percentage of my land tier....

What are you comparing with to say "load increased"? SL has grown 10x in the last year without, from my point of view, a big increase in the number of problems.


Five nines is _really_ expensive, though..

Exactly right. If they had the grid to themselves for two weeks, *nothing* would work afterwards. And given the usual 5 hour maintenance window usually runs about 8 hours, the two weeks would extend for... hmmm, I doubt it's a linear function. "Eternity," perhaps.

your cost effective solution?

MySQL is very robust as far a ROI goes.

Maybe your right, Maybe this should be a Banking Quality Asset Server.

but i dont need the extra fees - lol

My users consider 'system up but completely or mostly screwed' to be downtime. I don't think they hit 99% from that perspective.

LL has been open for to us residents for how long and they still fuck up royally, LL get it fucking right or you will lose all your precious money that we give you guys. I think LL no matter what they do how many weeks they use to fix this it wont help they can't do it right. Guys go back to school and learn a few things.

They're hiring.. If you think you can do better, apply.

there is java, ya know.

Are you threatening to boycot LL? Lol.. they have enough secure users to not worry abuot you hon

Okay, I'll bite. Don't let the brochure-ware fool you: every commercial DB is challenged by large scale distribution and real time performance. And they all have strange idiosyncracies that explode with scaling, so one learns on a per-platform basis where to kludge around the specific engine's limitations. It's not that one of the commercial DBs might not be marginally less kludgey, but that getting the current level of performance and reliability from it would take tens of man-years.

But all distributed DBs share one big vulnerability: the network. How many outages have we had due to connectivity to or within the colo facility? When the network has problems, distributed query processing tanks quickly, and trying to code around this is horrifically complex. Personally, I'd keep MySQL and spend the money on network capacity, reliability, and manageability.