December 26, 2002 at 11:14 pm
Hi all,
I wanted to shift tempdb from 0+1 to raid 0 array. I am not worrid about the transactions or rollback. if tempdb raid 0 array crashes, and I put new disk raid 0 array in place of crashed one will server start without any problems? Any other concerns except the transactions are losts.
Cheers..Prakash
Prakash
Prakash Heda
Lead DBA Team - www.sqlfeatures.com
Video sessions on Performance Tuning and SQL 2012 HA
December 27, 2002 at 5:04 am
No issues that I can see.
Andy
December 27, 2002 at 8:17 am
No problems that I know of....
December 27, 2002 at 9:45 am
Server might crash if you lose the array. If you're good with that, then no issues.
Steve Jones
January 1, 2003 at 12:47 am
Server crash is an important one.... Ok I assume if it happens then we need someone to repair raid array immidiatly... well that's a problem, availability of that person is a problem. Is it possible while restarting server after crash, i can change path for tempdb to some other working raid array???
Cheers...
Prakash
Prakash Heda
Lead DBA Team - www.sqlfeatures.com
Video sessions on Performance Tuning and SQL 2012 HA
January 1, 2003 at 10:49 am
I would say keep a spare drive handy and for sake of checking the drives condition swap it in the array with another periodically. As long as you can get the array back up quickly then RAID 0 is best. If however you are concerned with server stopping due to loss of tempdb use RAID 5 or RAID 1 since both will operate until you swap the drive and the tempdb will be available. RAID 10 is nice but costly for something that SQL will rebuild whenever restarted so data loss is all that occurrs in it.
January 3, 2003 at 3:02 am
Why have a spare disk sitting idle when you could stick it in as a mirror and deduce the risk in the first place ?
January 3, 2003 at 4:06 am
quote:
Why have a spare disk sitting idle when you could stick it in as a mirror and deduce the risk in the first place ?
Although I do agree with use mirror instead, I think this shows a basic problem with mentality of many DBAs and some System Administrators. Whether or not you need it you should have a replacement drive on hond ready to go at a moments notice. Sure it is idle but take my case in point. I have barked for years to have spares on hand but due to budget constraints this would not be approved. Earlier last year my point was taken (even thou they still haven't done it) as we lost a drive on our DATA array (RAID 5) and the vendor even thou getting it to us in 24 hours (I hated waiting) sent the wrong drive (oops), it took almost a week for us to get a replacement drive due to this mix up. As I explained during the time of the occurrance I do not accept responsibility for loss of data should another drive be lost during the time it was going to take for them to get us the drive. All in all unless management won't work with you, you should have a spare available at all times or one day you might find yourself in a pickle.
January 6, 2003 at 6:58 am
Would it be possible to use a hot spare in the RAID 0 configuration? I've used hot spares in RAID 1 and 5, and found the technology very useful when a production drive goes down.
January 10, 2003 at 2:48 am
I would DEFINITELY recommend having spares on-site. To argue on the grounds of budget is to be penny wise and pound foolish. It's like scrimping on your insurance, or fare dodging. Yes, you will get away with it a number of times, but when you get caught out the consequences don't bare thinking about.
We had a server go down and even though it was on 4 hour call-out it wasn't fixed for nearly 2 weeks!
The problem was that it was a 3 year old server and the manufacturer had trouble locating the parts, in addition, when the parts were finally located the replacement part was faulty.
Of course, this took some diagnosis as, having installed a new part the engineer assumed that that part was OK and that something else was at fault. Every component in the server, except the case and power-supply was replaced until the engineer found that the original replacement was faulty.
Fortunately I have a paranoid backup strategy so was able to switch stuff over to another server, but it meant some critical software was unavailable for the duration.
I would not like to put a cost on not having this software available but it exceeds by far the cost of keeping parts to hand.
A canny DBA would not only put (in writing) in the cost of buying the part, but also the cost of not having that part should it be needed. Said DBA would also add in the admin cost of having to run around like a headless chicken trying to acquire that part.
Once you put a monetary value on a level of risk then the decision makers can then make an informed decision. They may still make the wrong decision, but in the event of fallout you have a record in writing of having warned them.
January 18, 2003 at 2:12 am
Thanks to all for responses.
I have verified with microsoft that if I put tempdb on raid 0 and tempdb disk crash then sometime sql server will also not start as if we doing something in transaction and rollback starts and as no data is available in newly created tempdb so it may fail or take hell lot of time to recover.
Here no guarantee to recover.....
Agree guys?
Prakash
Prakash Heda
Lead DBA Team - www.sqlfeatures.com
Video sessions on Performance Tuning and SQL 2012 HA
June 23, 2022 at 9:56 am
This was removed by the editor as SPAM
Viewing 12 posts - 1 through 11 (of 11 total)
You must be logged in to reply to this topic. Login to reply