What is on Your Mind?

Eronysis -

OK. 350 servers, 6 people.

I assume you have root access to all 350. Have to so you can cover for anyone in the event of illness, vacation, etc.

But for the survey, would you respond 350? Or the 60 or so that you are personally handling?

I.E. Did you respond 50-100 or over 100?

I'd go for "over 100". If you see my first post in this thread, I said:
I have free rein only on the servers that I am building at any give moment. If I need to verify something on an existing production server (if, for example, I am building a server that needs to be cloned from an existing server), I login, take root - only if I must, check what I want to and logout. All of it in a matter of minutes.

Blowtorch -

I follow your logic.

But if you follow mine, then your answer would be just the servers you are currently building, plus the one more you access to check stuff.

I mean, sure, you technically have root access to all of them. But how often do you access any of them other than what you're building, plus that one extra server?

And I'd be willing to bet that the extra server is probably somewhere in the building where you are working.

Dave,

Very good point, I responded over 100. Albeit other than scripts running checking up on things via stored keys, I bet I only manually login to 40-50 per month. I see where you were getting the incredulity reaction from now.

Aaron

Another perspective...

When you have 2 or 3 servers, you can probably afford the time to rebuild each server from scratch several times a year.

When you have 2 or 3 hundred servers, you need to regularly monitor them - check statistics, clean up stale mount points, identify core dumps (and resolve the issue behind them), etc. Consistency is key. There's always something that needs attention. And because no system exists in a vacuum, there's always some variety of change. Business requirements increase (or decrease), legacy hardware needs to be replaced, new gear needs to be integrated...

If you let any one system get to the point where it requires severe maintenance, that time crowds out your regular maintenance, and you can easily end up on the wrong side of an avalanche of work related to sick servers.

In any given day, I may only touch 5 or 10 hosts. But I manage ALL of them.

Mostly by dial-in access. All SCO open server 5.0.5.

I am in a large disaster recovery shop. We have acres of systems of every flavor. I would say 90% of the time. 90% of them are idle.. unused.. root access available. If there not idle, customers are using and we cant touch it. Its amazing. Of course when they are idle.. they are not networked. So there is not a whole lot you can do with them other than play with the OS. But every industry flavor. Blade servers. 128 cpu superdomes. RS/6000. Sun. AMD/Intel.

And about 120 TB of SAN DASD that my team manages.

The thought of a uber-sneaker-net-seti@home monstrosity comes to mind.

At last rough count. between HP/Sun/RS6000 ~ 2500 CPUs not counting a farm of intel/amd based boxes.