Important Notice

[Stardragon]

Just to let you all know, the server will be down for maintenance from 10 pm tonight (June 20), until 10 am tomorrow (June 21). That's EST for all our overseas users =)

We apologize for any inconvenience.

[Halon50]

Any new features entering during this maintenance, or just general clean-up work?

[Jodie]

Not to be tooo bitchy, but. . ..

You realize that if the results servers are offline for 12hrs you folks will most likely have just killed all ~200 of my processors / 5.2M units/day. Do you have any idea how much work bringing that many processors back online will entail? It could be a week before I get that kind of time again.

I sure hope it's 1) worth it and 2) not a regular occurance.

Any chance you could please give us a maintanence window ahead of time next time if you can? Atleast that will let me bring them down in an orderly fashion and plan for the time to bring them back online...

Now I need to go turn-off my cell phone so the pages don't keep me up all night... Paint me an unhappy camper. :( :( :(

[RipItUp]

I'm missing something here .. why does not being able to connect kill the processes ?

Surely they will just buffer until it can conenct again if you are using the -df function ? Or are you running differently to mine ?

At least being a 13 year old girl you'll have gazelle type legs to dash about popping them all back on !!!

:teasing:


Regards

Andy

[IronBits]

Stats Server is online :D

[Jodie]

Originally posted by RipItUp
I'm missing something here .. why does not being able to connect kill the processes ?

Surely they will just buffer until it can conenct again if you are using the -df function ? Or are you running differently to mine ?

At least being a 13 year old girl you'll have gazelle type legs to dash about popping them all back on !!!

:teasing:


Regards

Andy

I don't run the -df flag. The smallest pipe any of them are running on is a 3Mbit bonded pair of T1's. The latest round are on a redundant 255Mbit pipe. I've never seen a reason to "nonet" them...

[bubbadog]

Originally posted by Jodie


I don't run the -df flag. The smallest pipe any of them are running on is a 3Mbit bonded pair of T1's. The latest round are on a redundant 255Mbit pipe. I've never seen a reason to "nonet" them...

In your postition, I would run the -df flag anyway. Servers are not infallible and things happen. Sometimes these things are scheduled; sometimes they aren't. I can understand your reasoning for not doing it; I'm just pretty cynical/paranoid about the fallibility of things (Yeah, I'm the poster girl for Murphy's Law :D)

[Richard Clyne]

Originally posted by Jodie


I don't run the -df flag. The smallest pipe any of them are running on is a 3Mbit bonded pair of T1's. The latest round are on a redundant 255Mbit pipe. I've never seen a reason to "nonet" them...

The -df flag is not a nonet flag. It increases the available buffers from 6 to 1,000.

This simply means that instead of the client stopping after completing six sets of filestructures it will continue on folding until about 1,000 sets of filestructures have been generated.

This should cover most outages I would assume.

[Moogie]

Thanks for letting us know Stardragon. I know these things have to occur from time to time. :)

Richard..good explanation. You beat me to it again. :)

Happy crunching all!

:cheers:

[Starfish]

Originally posted by Jodie
[B]Not to be tooo bitchy, but. . ..

You realize that if the results servers are offline for 12hrs you folks will most likely have just killed all ~200 of my processors / 5.2M units/day. Do you have any idea how much work bringing that many processors back online will entail? It could be a week before I get that kind of time again.


~200 processors :shocked: :notworthy

Amazing... I think you've brought the dream I had before the drawing of a large national lottery into reality ;)

Unfortunately my lottery numbers were not the good ones, else I knew what to do as well :thumbs:

[Jodie]

Thanks for the explanation, all! I will migrate to the -df flag. I guess I don't understand why it's even an option, if there's no detriment to running it and nothing but upside - why not build it in as 'standard equipment'?

---

You can take back your :notworthy smiley - only a hundred or so belong to me. The rest are a couple of client clusters that we need to burn-in. Fortunately the burn-in and porting times are relatively long (2-3mo) and the number through the doors is pretty steady, so I should be able to maintain atleast 200 reasonably fast (1.4G/512 P3 Server, Xeon P4-2.2 and AMD 1800+/1900+) processors for the forseeable future... I'm going to start bringing up a few of my old SGIs, I'm curious to see how the Challenge, O2 and Octanes run...

[Starfish]

I'll keep the :notworthy smiley because I think it's great that people bring so much cpu power to a good cause :)

Cheers :cheers: :D

[Aegion]

Originally posted by Jodie
Thanks for the explanation, all! I will migrate to the -df flag. I guess I don't understand why it's even an option, if there's no detriment to running it and nothing but upside - why not build it in as 'standard equipment'?

The reason is that a large amount of caching can eat up alot of hard drive space. Particularly in an office enviroment, this can be a significant issue and potentially lead to the computer users objecting to an authorized individual borging their machines for Distributed Folding. This is why a limited of six buffer sets are set by default for the client.

[Jodie]

Thanks, Starfish!

And thank you Aegion, that makes sense to me. My cluster-based systems are network-booted and live on, in one case, a 2.5TByte raid and in the other case, a .5Tbyte raid. So disk storage isn't an issue for me personally on those... But it does make sense. I'll change to the -df option as soon as I have time to start bringing them back to life. M2kGuy - make as big a run as you can QUICK! You can take back the #1 spot for a week... ;)

[Starfish]

Originally posted by Jodie
Thanks, Starfish!

And thank you Aegion, that makes sense to me. My cluster-based systems are network-booted and live on, in one case, a 2.5TByte raid and in the other case, a .5Tbyte raid. So disk storage isn't an issue for me personally on those... But it does make sense. I'll change to the -df option as soon as I have time to start bringing them back to life. M2kGuy - make as big a run as you can QUICK! You can take back the #1 spot for a week... ;)

But I bet you don't want to upload 3Tbytes of structures :rotfl: :rotfl:

[Scoofy12]

Originally posted by Starfish


But I bet you don't want to upload 3Tbytes of structures :rotfl: :rotfl:

:bs: Heck, she can upload em to me.... as much as she can pack over my 1mbit pipe... ill get a new HD and see that they end up in the right place (my account:jester: )

[Jodie]

Office network wouldn't be all that bad... Home network at 3Mbits might suck a bit... But it's the smallish raid anyway...

[Jodie]

Don't tell anybody - I wasn't 'feeling well' and 'went home' to 'lay down' for a couple hours. ;)

[Starfish]

Originally posted by Jodie
Don't tell anybody - I wasn't 'feeling well' and 'went home' to 'lay down' for a couple hours. ;)

:rotfl: :jester: :rotfl: