Main grid up. preview still down.

The main grid's been up for hours, but the preview grid is still down.

Is it out for the count, or will it be reopening?

I answered earlier here when you asked.

/139/9e/105831/1.html

I don't think I did ask earlier. Are you possibly getting me mixed up with someone else?

when will newbies be able to test the preview?

This is happening far too often! What kind of shitty hardware is running the database server anyway?

Doug Linden has just posted...

/3/ef/106215/1.html#post1034158

Preview Grid is back up!

Good news indeed, however...



Not happy about this. I wanted my shadows. T.T

Yeah, me too... and then time we'll pass and we'll get 'em back--and by then there'll be even more improvements on the horizon. Always such a road......

OK, this is freakin' ridiculous. For this to even occur is just stupid in this day and age of RAID and, supposedly, backups. What kind of idiot-braindead colocation is LL using that has THIS much trouble with keeping its hard drives and databases in order?? I mean it's pathetic! It's shit like this that does NOT make me want to even bother testing the preview (beside the fact of having to download 20MB updates every few days). I didn't have time to get copies of all the scripts I had created (though I had some). And I still don't appreciate having all my stuff in Morris just deleted for no reason given while EVERYONE ELSE'S STUFF REMAINED! If this is how you treat someone who bothered to take the time to submit almost *40* bugs, I simply won't bother anymore.

It's a preview grid, Eep. It isn't the main grid. Would you expect the same kind of strain on a cluster of servers that shows off alpha or betaware while a stable release is present? They aren't even on the same level as production hardware. I'm not all that pleased about the databases failing, but it happens. Hard drives break down eventually. "Life sucks, buy a fucking helmet."

And 'twas my understanding that to even get on SL you need some form of high speed connection. Are you saying you can't get a 20 meg download in less than a minute? Go get a drink, read a web comic, ANYTHING. I go see my mom off for the day and by the time I get back up to my computer the patch is downloaded.

Again, have you not heard of RAID, Feynt? It's mindless enough to install so there's really no reason a hard drive failure should do anything than simply prompt a replacement drive installation on the failed drive. Duh. It ain't freakin' theoretical physics!

You're missing the point (as usual), Feynt. SL wastes bandwidth like there's no tomorrow. It's not just having to redownload the client every time that's annoying, but having to reinstall it, which recreates the deleted "Second Life Preview" (as I renamed it from the stupid "Second Life Siva" program file group and the desktop icon. Plus I have to overwrite colors.ini, keys.ini, and viewerart.ini each time, which is a pain (even with a batch file). It's about efficiency, Feynt. http://tnlc.com/eep/sl/ to learn more ways SL is bandwidth-ineffcient/-wasteful.

Yes, I've heard of RAID. My friend has a lovely 1.25 terabyte drive array he likes to bring up whenever mentioning his anime collection. And I'm sure you know that if key drives die, for whatever reason, your data is STILL screwed. Sure you can recover, but weigh rebuilding a partially erased database against taking a snapshot of a complete, new database and replacing the old one. Sounds a lot more appealing doesn't it?


As for wasted bandwidth, I won't argue with you. 20 meg downloads instead of a small patch program and minor downloads that patch existing files IS wasteful. But it solves the issues inherent with minor patching as well, which can in its worst cases lead to needing to redownload the whole thing anyways. More wasted time versus and automatic reinstall. Or if you just download the changed files, there might be compatability issues unforeseen by the developer who always have updated files. It's true laziness and lack of foresight at its best, but it works and it's only really in the preview grid that you get this major waste of bandwidth (how often do patches come out that require a download on the main grid?).

Caching, yes, definantly needs an overhaul as well. I feel like I might as well turn it off some days. I'd rate that high on the list of "things to fix" for the next major release. But it's easy to say "it's broke, fix it!" and harder to say how.

No. It sounds like you don't have your facts straight about RAID. A MIRRORED RAID setup doesn't HAVE a "key" drive since it's 2 drives per same data so if one goes, there's a FULL backup ready to replace it.

4 times in the first week I joined in 12/04. More often than you might think (or care to admit).

Yes, well, apparently it's easier for LL to add new features instead of fixing bugs too, so...

The only thing RAID would protect against is a single drive failure. How do you know this was the cause of the database corruption? I agree backups are sensible, but I can see why they are not made so frequently on an alpha/preview system.

Right...and the likelihood that BOTH drives will fail at the same time is VERY slim.

I don't; it's just one possibility. But SL's database seems to get corrupt FAR too often, so perhaps they should dump MySQL and go with a REAL database system.

They should be made at least daily on ANY system (and even more frequently on time-critical systems).

Depends on the RAID; if you have a hot spare in place and can configure for autorebuild, then you are vulnerable to a two-drive failure for the time it takes to rebuild the drive.

Ultimately, it boils down to cost & time vs. inconvenience. Obviously LL has decided that it is acceptable to take hits on the *preview* grid. Now, if we were talking about the main grid....

That's exactly my point, Seraph. It's the preview grid, anything that DOES happen there (database wise) is temporary. Nothing you make stays with you or goes to join you in the main grid when the update goes live, so why make a super redundant setup that costs as much as production when you're just going to scrap the data anyways?

I think though, Eep, that it is you who are confused and without all the facts as far as RAID goes. You seem to be thinking in terms of RAID 1 format, where each drive is a mirror of the others. There are in fact 5 other levels (RAID 0, arguably not a RAID format at all, and 2-5) which do NOT mirror data like this, but instead use a set of drives to store parity information about the files stored on the other drives. Depending on the level this can mean an extreme speed increase (read access split across 4 drives for instance), and the parity drive will ensure that data that's read will be correct. Should any single drive be damaged, the information from the other drives can (emphasis, CAN) be used to rebuild the lost data. However if the parity drive AND one of the other drives is damaged, you suffer an irreparable loss of data.

RAID 5 suffers from this the least, as the parity data is spread across all drives to mitigate read access to the actual data and the parity data. This allows multiple file requests at once, gives a good increase to access speed, and should one drive fail it's possible to recover much of the data because in theory only a fraction of the parity data is lost.

Regardless though, if you suffer more than a 1/4 loss of your storage drives, it becomes impossible to recover data. All it takes is a power spike, a malfunctioning drive controller, or some dork who got too close with a cup of coffee (and I've watched that happen. Don't say it's impossible).

Addendum:
Actually, RAID 0 doesn't use parity data, it merely spreads file data across the all drives. Lose a drive and you lose your data with no chance of recovery.

If that's the mentality, I won't bother testing previews.

I know enough about RAID to not bother with anything but RAID1, if at all. Any other RAID format is just as risky as running non-RAID so they're really just not worth it--not even for the speed increase. For LL to be running anything other than RAID1 (on either preview or main grids) is just foolish and stupid.

I disagree, because as you had said the likelihood of a failure of such a magnitude that you'd lose a critical amount of data is slim. RAID 5 really is a safe enough system, and allows you enough security to rebuild what data you might have lost. Combined with periodic backups of data to a seperate storage system (which itself would be the RAID 1 to the RAID 5 system they probably use).

Of course on a production server you should be swapping hard drives every three to four years regardless of wear. A testing server doesn't need that kind of replacing because.... it's a testing server! It gets hand-me-down equipment you phased out of the production servers.


And what's with this remark? You KNEW going into this that nothing you make will remain with you once the update went live. So why did you bother making stuff you'd get attached to in the preview? Why can't you remake it?

Aside from it being stupid that preview grid-made stuff can't be brought into the main grid (never have liked the idea of separate grids anyway), yes, I can remake everything but it's just time-consuming and I didn't get a chance to get all my scripts and/or prim params for everything I made (a lot). The point is: had LL had a better backup/database system, there's no reason the database WOULD have gotten corrupt, which leads me to believe whoever's running the preview grid's database, at least, is incompetent (and it might explain the Database Warehouse Engineer and/or part-time/intern operations positions on LL's employment page).

Well since the positions are there, and you seem to know what to do to make it better, why haven't you applied? >

The preview grid is there to test everything. EVERYTHING. Buying stuff included. You get a hefty amount of free money and there are several vendors from the main grid hanging around in the preview grid as well. If you could bring everything from the preview grid into the main grid, that would make all that stuff worthless. Just wait for a preview and it's yours for free. It would also prevent people from being so open about handing out their cool stuff. I've gotten free capes and tentacle monsters to play with, and handed out my torches to several people (crappy and hastily done as they are). Last preview I'm told one of the Lusk folks was handing out free taur avatars to everyone he met. You don't get that kind of friendliness on the main grid.

I don't code.

Um, testing buying things hardly takes an entire grid. Test vendors could simply work in "play" money or, even more simply, just refund the money after the test--and only offer simple, no-real-value content (like duplicates of the "Library" inventory folders).

You can't test the buying system without paying money. And while yes, you could have it refunded, you still need money to begin with. Hence, free money. But buy testing is only part of the process of the preview. The point is with everything being temporary there's no value placed on anything made there. Everyone freely gives others stuff to learn from (like neat wavy hair, or lovely light accentuated landscapes), and in the end you carry away more valuable stuff than things you've made: You get friends and ideas.

Money isn't necessary to test the buying system, really. You already agree it can be refunded so there's no reason the vendor couldn't simply GIVE you the money initially to test with ONLY on that vendor.

Ideas and friends are still possible in the main grid.

But you have no reason to restrain your generosity on the preview grid, you can hand out your cool widgets with impunity and get to know people. I've met a few people on the preview that are now friends in the main grid because we shared a few neat ideas.