Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Problem: Our server becomes very very slow (near freezing) if I copy a file from any directory on C: and D: to any other directory. In short It becomes very slow if I touch the hard disk. Currently I have turned Oracle off (It's not our main database) and things are working but very unstable and I can not touch anything.

What I have found:

What we have there: Windows 2003 R2 Enterprise x64 SP2 (Uptodate), Oracle 11g (Currently all of it's services are down; so it is not actually working; but before this, it was being updated by our main server via Oracle Streams), ~10 .NET applications (3 of them make big log files ~5 GB/day) and 20 ASP.NET websites and services, 2 of out applications talk using MSMQ (which is very important to us and it's performance drops to 10 messages/sec when we where passing near 100 messages/sec before) we process near 3,000,000 messages/day.

Applications (and their logs) are on D:, Oracle data is on E:.

  1. C: Total Size 44 GB, 12.5 GB free
  2. D: Total Size 295 GB, 35.5 GB free
  3. E: Total Size 1.36 TB, 282 GB free

Hardware:

  • Intel Xeon XPU 2.00 GH, 4 cores,
  • 48 GB RAM,
  • 6 * Fujitsu MAX3073RC 73GB 15000 RPM 16MB Cache Serial Attached SCSI (SAS) RAID 5 (C: & D:),
  • 2 * Seagate Barracuda 7200 1.5 TB 7200RPM SATA 3Gb/s 32MB Cache RAID 1 (E:)

Our admin is unreachable until next week; and I am not the admin :\

share|improve this question
    
Server 2003 R2 "Up To Date"? Is that a joke - that product is 3 generations behind. And on a VERY slow disc subsystem - and you wonder why performance is bad when you copy? I would assume disc response times to close 3 digit in miliseconds, isntead of being low one. –  TomTom Jan 15 '13 at 21:32
2  
@TomTom - to be fair, Server 2003 can be up to date as in it has all its patches applied. Also, one would not expect a file copy to make the entire server virtually unusable. And yes, the 7200RPM disks are slow for a server, but it's not "very" slow. –  Mark Henderson Jan 15 '13 at 22:01
    
@MarkHenderson They are VERY slow. Seriously. As slow as 720 RPM discs are in a high IO setup. Yes, they DO die - whoever was setting this up was not counting 1+1. The discs have around 150 to 200 IOPS only, and all database side operations are constantly issuing interrupting flushes. They are totally ok as a file store, they totally suck for what they are used here, as it seems. Same with the RAID 5 - the IO budget of them is low especially on the write side. –  TomTom Jan 15 '13 at 22:22
    
@TomTom - you are correct, I somehow missed the fact that he's running about 1Tb of an oracle database on them. –  Mark Henderson Jan 15 '13 at 22:29
    
@MarkHenderson Yes. I run a 800gb SQL Server and it has 16 Velociraptors (300gb version) in a RAID 10 and 2 SSD for the logs + small SSD For the OS. (and soon gets a 7805Q and large SSD For caching). THAT is more what you need to run a high performance database setup (and even that has IO problems, thus the SSD cache coming). A Mirrored 1tb ulto low IOPS thing things - oh my god. –  TomTom Jan 16 '13 at 5:59

1 Answer 1

Use procmon to check your disk queue lengths when this happens, as well as looking for bad disk sectors or trouble with that RAID5 array (which is what my experience tells me is probably the problem).

I should add that I had what sounds to be the exact same problem with an ESXi host a few months back, and it turned out that the RAID5 array was degraded (had one disk fail), and encountered a URE (unrecoverable read error) that brought disk activity to a standstill when certain files were accessed.

As an aside, that configuration seems all kinds of... ridiculous. You pay a ridiculous price for Oracle, and put the high-performance-for-an-obscene-price-tag Oracle database on 7200 RPM SATA disks, to make sure that's wasted money... and then pay a premium for 15k RPM SAS drives, that you then throw into RAID5 and run your OS, applications and logging from, to make sure that's wasted money too?

Sounds to me like you'd benefit from having someone who knows what he's doing come in and set up your disks and array properly. Maybe on an OS that isn't on the verge of going EoS, while they're at it...

share|improve this answer
    
+1. Add a "current up to date" 2003 R2 and the question is - is that on purpose? Sabotage on the company? This MUST be a joke. –  TomTom Jan 15 '13 at 21:33
    
I would add a decent setup, get and plug in some SSD - especially with all the transactional stuff on it, the SSD will make this server possibly scream in performance, compared to what is there right now. –  TomTom Jan 15 '13 at 21:45

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.