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

With this new update the sims will be more stable? cause mine infernal ville residencial dont stop to crash since the last update was maded. Its use to crash 5 or more time per day. I hope this new updates arrives near of the perfection.

There is absolutely no excuse for the incompetance of scheduling appointments, events etc that clash with Linden Labs maintenance schedule. There is sufficient notice of any of the scheduled maintenance tasks that Linden Labs need to carry out and are listed on the Grid Status page on the right side panel. You can even add these to your Google Calendar. I check this page on a daily basis and have the link on my Firefox Bookmarks Toolbar as well as set in my Freeview TV in-world.

As for the business types whinging and exagerating about a miniscule five minutes downtime to improve the Grid. wOOt, big deal. Loosen your tie. I cuts off the blood supply to the brain and ruins the intelligence. I suggest you concentrate on the viability of your virtual business and leave enterprise level computing to those of us who are experienced and qualified. Just because you can gap your own spark plugs does not make you an auto mechanic.

There is one problem though with the blue dialog giving notice that the region is restarting. I think that we only need ONE warning (5 minutes from memory) and not several. In the last minute there are several, every 10 seconds approximately. These actually jam up the viewer and as these dialogs have focus() they prevent you from logging out or teleporting. Seeing as most of us know how long 5 minutes takes (I have a watch and a system clock in the system tray) subsequent notices are superfluous.

Why in earth would you bother to do this? For what reason?

Are you suggesting that all residents clear all objects and large build castles from their parcels before a rolling restart?

Yes i do on complex builds because in the past prims have gone missing

I have never had a single prim go missing after a restart - ever!

Well lucky you

Wow, almost on time.
The update service for "Sarge" had been discontinued in March 2008. That means, you are about 11 months behind with security updates (and there were some serious issues with Debian's version of OpenSSL last year! -- just to mention one), unless you backported or compiled a huge chunk of the system and tools manually (which I seriously doubt).

But the good news from my experience is that the "dist-upgrade" for about 300 servers last year worked seamlessly (apart from a small problem with the SASL libraries, but I doubt you use that on the simulators at all).

So from that point of view you made a good choice with Debian

I'll keep my fingers crossed that the updates work without problems! I know that this isn't a fun task to do -- especially not for your amount of servers.

* Normally during rolling restarts, we lock regions to hosts. Most of the time when a region restarts, it will migrate to a different host, but they don't do that during rolling restarts. However, this rolling restart is different; because hosts are not just having a Second Life software upgrade, but are being completely wiped and reinstalled, we won't have the freedom to lock regions to hosts.

* Normally during rolling restarts, we lock regions to hosts. Most of the time when a region restarts, it will migrate to a different host, but they don't do that during rolling restarts. However, this rolling restart is different; because hosts are not just having a Second Life software upgrade, but are being completely wiped and reinstalled, we won't have the freedom to lock regions to hosts.

Do I understand, my region will change its server every time I restart it? Performance was the poorest after the last two restarts - this would be the solution to this.

There is a fix for this ... much like working on an extensive document in Word on in a spreadsheet ... save often!

What Prospero was saying was that during the rolling restart your region would likely change its server.

Thanks for giving me enough warning to send a notice to my group. I don't usually announce rolling restarts there since I don't know which of my regions will be affected over the typical three days, but this sounds potentially severe enough to unnerve my renters.

By the way, how does one volunteer to be in the pilot rolls?

That's not guaranteed, but it is possible. My understanding is that, when you initiate a restart, your region will boot up up on the next available server, which may or may not be the one it was running on before the restart.

Things rezzed out in regions should survive a rolling restart. That being said, rarely, yes, changes made within the last hour before a restart do get lost when the region restarts. Regions save their contents every hour, so that we have backups, and so that we don't lose more than that. They also save their contents on shutdown. However, if something goes wrong on shutdown (which, in the vast majority of cases, it does not), then we will have to revert to the previous autosave.

In any event, when you're building, it's always prudent to keep regular copies, just as it is with anything else you do on computers. I know myself, both building in Second Life and with other things I've done on computers (writing, programming, etc.), I've lost stuff due to my own missteps (accidental deletion) much more often than system problems... so, there are always good reasons to keep copies of what you're working on.

During a regular roll restart, you get 5min of warning before the restart hits a given region. With this etchification rolling restart, you will have 8min of warning.

i smell a big screw-up coming !!!!

This rolling restart should really not be much more severe than any other rolling restart. It will continue over several days, and, yes, some regions will get hit more than once, but other than that people will mostly not notice it's going on.

Re: the pilot rolls, email me at [email]prospero@lindenlab.com[/email] if you want to have your regions included.

Prospero: Here is my problem with even the minimally disruptive rolling restarts: predictability. I have long thought of starting a business in SL that would involve bringing in new users from RL small/medium businesses for small group meetings, and the platform is stable enough to do this in an unambitious way. However, if my sim were to go down for 5-10 minutes in the middle of a meeting that a client has paid (and trained) for, I'll never see that customer again.

*shrugs* It's a matter of priorities for LL. If you want RL business folks in SL, then just like you have sims volunteer to be pilots, I would think you could have sims subscribe (for a price), to only be upgraded on a schedule. You have the het grid - use it!
.

I'll take that under advisement. I'll definitely be informing my estate managers, at least, of the potentially unusual nature of this restart and refer them to your post rather than confuse the issue with second-hand notices. Come to think of it, I should get back to doing that with regular restarts, too. Oy. I take a week off and start getting sloppy....

Thanks for the info. I didn't want to contact you directly if there was a more formal route I should be taking, e.g. a support ticket.

Predictability? Aren't rolling restarts announced?. Why you can't reschedule a meeting?

Yes, they are. Grid restarts for non-pilot sims are typically now done over two 12-hour periods; this one will require 3 12-hour slots. If you want to avoid restart risk, that's a lot of time to keep your business offline.

Of the last 3 gridwide restarts, two were given with 1.5-2 days warning, one got a few hours warning. It can be difficult to reschedule widely dispersed, perhaps international, meetings with a day or two's warning, let alone a few hours. If the service you are charging for requires rescheduling at all, it looks bad. If the rescheduling has to be done on short notice, for many it would be just as well to cancel the virtual meeting and hold it using Webex at the originally scheduled time.

I don't see why businesses, for a fee, couldn't schedule most restarts for a single half-day slot a few days after the rest of the grid restarts. That would effectively stretch out the warning time, and lessen the time that the business had to be idle.

Of course there will still be the occasional, unavoidable, urgent maintenance restart. And of course LL would prefer that the grid all run the same version. But most restarts nowadays seem to be less urgent.
.

Why can't businesses have conference rooms on a couple of different sims, if they're concerned about not being able to get into the sim when the meeting's due to happen?

In fact that could be a decent business model, renting out conference facilities.

hm An idea for a few of these comments. Get your sim online at the beta grid and have it setup exactly as it is on the main. I know the problem could be of people not being able to access it because of age but I think if maybe a tentant (spelling) agreement was signed with ll that could be over come. dual purpose area's so main is affected you jump to beta and vice versa. just my thoughts

The answer is that LL are pushing SL as a BUSINESS PLATFORM. That means that they want companies to come into SL and use it as a meeting room,. advertising platform, etc.

If the platform ( note platform, not game) is going to be unavailable for some undefined minutes at some time during three days, then the platform is completely unreliable for serious use during the complete three days.

This means that SL is no use whatsoever for serious use during these times.

For building and other game related activities this is not a real problem , merely an inconvenience, for a company, it can be a very real expense.

This is the point that Atashi is trying to get over and which Prospero Linden is completely missing.




Good luck with the update anyway

Hi Prospero, good luck with the Etchification of agni.

BTW, wouldn't it mitigate the possibility of a region being restarted more than once if you restarted all servers currently without regions at the start of the process? Just curious.

For those of you wanting more "predicability", perhaps you should consider volunteering your regions into the 'pilot roll program'. That way you only have to worry about one short period of time (if all goes well) rather than the three days of perceived uncertainty. Of course, there may be other risks being in the pilot group, However, your restarts would be more predictable anyway.

Also, there are huge blocks of time on RR days that are free of RRs. For example, Tuesdays restarts are from 8am-10am & 8pm-8am. This leaves 10am-8pm Tuesday, 8am-8pm Wednesday, etc. for meetings. Of course I understand that people from all time zones are in SL and these blocks of time may not be convenient for some.

Please remember folks, LL doesn't do these restarts just to annoy you. The restarts are done to; install fixes to software that we will all, hopefully, benefit from; to upgrade hardware or software that will likewise benefit our SL experience. So, cut them a little slack and use your imagination and creativity in your meeting arrangements so that the scheduled restarts don't interfere with them.

Take care all

Without private grids behind corporate firewalls, or making the whole SL protocol firewall-friendly so it can be used through a reasonably secure firewall, it really doesn't matter.