Discussion Locked

Server Reboot 2003 Best Practices

0 helpful
2 replies
Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.

We have a debate going.

I say: set the mission critical server W2003 to automatically reboot every monday morning at 3:00 AM to free up resources and place system in optimal state for upcoming work week - thus increasing availability.

The opposition says: this places excessive stress on the hard drives (raid 5) and other components, ie. fans, power supplies, etc. and will cause premature failure of components.

I have tons of stats that disk drive testing incorporates stress testing, start-stop, etc. when determining Mean Time To Failure (MTTF)so I think this arguement is empty.

I cannot find anything from a Best Practices view point that is backed by research. You can bet someone has the stats but I just can't find em. Server is W2003, dual quad cores, 16GB Ram, Raid-5, high speed SATA drives 15K rpm. I am looking for info on general, everyday, business conditions.

Thanks

2 replies

by Member 20444064

I have been asking questions like that for years with no real answers.

Here I reboot my 1/NT, 3/2000 & 3/2003 servers & 1/SBS 2003 server every week.
I run Citrix/Terminal Services on 2 of my 2003 servers. If I go more than two weeks my president/CEO starts complaining things are running slow remotely. I have 50 users who can tell when I reboot because they comment that the system seems to running faster after.
Just running the system puts stress on the hardware. I think not rebooting regulary could cause more problems than rebooting. I know from turning on a computer thats been off for a few months can be tricky and have problematic results.
We all know windows has a memory leak so a reboot is a must. Whether weekly or monthly I agree it needs to be done. There will always be that someone who says I reboot yearly. I say how much does the server really get used. My are used 7 days a week with 5 days of 16 hr days.
I reboot my desktop only when it needs it. (XPPRO)

The question has always been there "Do I turn my computer off every night or leave it on?" I say depends on how much you use it. I use mine (Home) every day so I leave them on. Except for the monitor, though LCDs do use less power, I used to turn off my 22" CRT after I was done.

The only thing I question is the automatic reboot. I have had servers not come back after a reboot. If you do it automatically that means you must be there in the morning if something doesn't work. We start at 6am & I don't come in until 0830. Thats why I reboot on Saturday or Sunday night.
My 75cents worth.

by furquhart

I am limited to a narrow window for reboot which is between 2:00 AM - 3:30 AM on Monday morning only. That's the only morning we don't perform a backup. As for a server not comming back up...We have a 24/hr in-patient unit that must document patient care. If the server doesn't come back up, the nurses detonate a load of C-4 hidden in my on-call cell phone as soon as I flip the phone open.

MTTF (Mean Time To Failure) on a drive is usually 1.4 million hours. That's lab time. In a recent study of 70,000 drives in real world use you divide the 1.4 million (lab time) by 14 and that give you real-world MTTF of 100,000 hours. Divide this by 24 = 4,166 days MTTF or 11.4 years MTTF. Laboratory MTTF testing includes stress testing, power on/off,sudden power loss,etc.

I think the amount of degredation from re-booting is minimal and the benefits are maximum. This is a Citrix server. I estimate the Business Interruption at about $1,000/hr when it crashes.

I think my Network Admin is short a packet or two in his thinking.

I've got references for my stats if anybody wants em.

Any comments?

Phred
DBA

BTW I think your statement is very astute, "I think not rebooting regularly could cause more problems then rebooting."

I'm adding some stuff to my post. MTTF is a statistical method of forecasting failure. It does not mean a drive will last 11 years. If MTTF is 1.4 million hours the statistic means if you took 114 drives and ran them for one year you could expect 1 drive to fail during that time.

The above is a lousy paraphrasing of article by Adrian Kingsley-Hughes, March 2007, http://blogs.zdnet.com/hardware/?p310, Making Sense of "Mean Time to Failure."

My apologies to Adrian.

An excellent article is from the Parallel Data Laboratory at Carnegie Mellon University. "Disk Failures in the Real World" Bianca Schroeder, Garth Gibson, September 2006.

This discussion is closed to new replies.