Sure. If you've got, say, 50% of your servers powered off, and you make a config change, do you have an automated way of powering on the other servers so that they pick up the change?
Or your configuration changed while the backup was powered down.
I used to have a BSD firewall where the backup was offline. Managing firewall changes were "fun". Quite a few years later, I can see about a dozen better ways to do it, but there you go
I concur. I work in the financial services industry and have clients in four continents. I don't really have a maintenance window. I've got a few hours on Sunday morning where I can't screw anything up too bad.
I think the problem is that we've gradually gone from cuneiform (lifespan undetermined. several thousand years so far) to blank CDs, which I can't seem to make last a few months without something chipping the cheap foil off the back
How long do you suppose you have to build data centers before Google hires you and tells you to go nuts? That's amazing, and I'd love to just walk around in it asking the guy questions who designed it.
I can't imagine how much electrical engineering, physics, and IT years of experience were required to put that thing together.
Anyone get to work in something approaching this level?
I asked about trade magazines and got a few responses, and I'm always looking for more reading material, so I'm interested in hearing responses too.
And for the record, I like Lightwave, Dr Dobbs, Network World, Storage Magazine, and I'll be writing soon for Simple Talk Exchange, so you should subscribe to that, too;-)
Really? I'd never thought of that, but I suppose it makes sense. Does the controller actually remap all of the individual blocks, or how does that work?
Interesting. I didn't notice the date earlier. It actually came across the RSS stream earlier from ZDNet. Even if it is old, it's an interesting topic that, if true, will have dire consequences before too long.
It really only deals with SATA drive (SAS probably has lower failure rates) and it only becomes a statistical issue with mammoth amounts of data (the amount quoted in the article is 1 data read error per 14TB)
Slashdot Mirror
Sure. If you've got, say, 50% of your servers powered off, and you make a config change, do you have an automated way of powering on the other servers so that they pick up the change?
I don't have the numbers in front of me, but failures are supposed to increase during start up of drives and power supplies.
Or your configuration changed while the backup was powered down.
I used to have a BSD firewall where the backup was offline. Managing firewall changes were "fun". Quite a few years later, I can see about a dozen better ways to do it, but there you go
I concur. I work in the financial services industry and have clients in four continents. I don't really have a maintenance window. I've got a few hours on Sunday morning where I can't screw anything up too bad.
Must be nice to be able to load balance by datacenter as opposed to physical (or virtual) machine.
What magic are you using to speak to us? Admit it, you used Google Translate!
Wait, you know of a place where sysadmins are paid overtime? Or even by the hour?
Out of professional curiosity (i.e. I'm going to steal your answer if it makes sense)
How do you manage config changes to shut-down VMs?
I didn't know my users posted on slashdot...
Or when config changes aren't replicated fully because of unpowered servers
I think the problem is that we've gradually gone from cuneiform (lifespan undetermined. several thousand years so far) to blank CDs, which I can't seem to make last a few months without something chipping the cheap foil off the back
haha Busted!
So seriously, can I come play?
I was wondering what made this illegal. My bandwidth providers do this all the damned time.
Nice, thanks a lot!
How long do you suppose you have to build data centers before Google hires you and tells you to go nuts? That's amazing, and I'd love to just walk around in it asking the guy questions who designed it.
I can't imagine how much electrical engineering, physics, and IT years of experience were required to put that thing together.
Anyone get to work in something approaching this level?
The problem with Libraries of Congress is that it's not a static measurement. You've got to have Libraries of Congress adjusted for inflation
I asked about trade magazines and got a few responses, and I'm always looking for more reading material, so I'm interested in hearing responses too.
And for the record, I like Lightwave, Dr Dobbs, Network World, Storage Magazine, and I'll be writing soon for Simple Talk Exchange, so you should subscribe to that, too ;-)
[CITATION NEEDED]
No, I'm really hoping you're right, but where did you find those numbers?
Really? I'd never thought of that, but I suppose it makes sense. Does the controller actually remap all of the individual blocks, or how does that work?
I've thought about using RAID 5 as the bottom level of some other level, but I can't convince my self that it would be slow as molasses.
As long as there's some recent copy of the data that doesn't rely on the controller not puking all over it's drives
Interesting. I didn't notice the date earlier. It actually came across the RSS stream earlier from ZDNet. Even if it is old, it's an interesting topic that, if true, will have dire consequences before too long.
It really only deals with SATA drive (SAS probably has lower failure rates) and it only becomes a statistical issue with mammoth amounts of data (the amount quoted in the article is 1 data read error per 14TB)
Reminds me of the definition of a lie:
A poor substitute for the truth, but the only one discovered to date
RAID-01? ;-)