Disk Arrays
cancel
Showing results for 
Search instead for 
Did you mean: 

array expand times

Chuck Parrott
Occasional Visitor

array expand times

I'm nursing along a 7 yr old ML350 server with 2000 and SQL on it for a few months until our software upgrade is ocmpleted. We were seeing some i/o performance lags that prompted us to buy an additional 73.6 GB drive to expand our 2nd logical volume that had 3 73GB drives prior.

How long will this process take? We're 24 hours into it with 4% completed. By that measure, it's going to take apporx. 20-25 days.

The controller's expand and rebuild priority is currently set to low. I've read setting it to high will help, but this is a heavy live production environment and I am concerned about user impact.
8 REPLIES

Re: array expand times

I would just let the Rebuild Process run..
Also would make sure i have the Necessary Data Backed up..
Once done i would certainly update the Firmware on the server and the Disk to make sure all in the same level and uptodate.

Regards,
I am an HP employee
Steven Clementi
Honored Contributor

Re: array expand times

does the server work hard on the weekends?
Steven Clementi
HP Master ASE, Storage and Clustering
MCSE (NT 4.0, W2K, W2K3)
VCP (ESX2, Vi3, vSphere4, vSphere5)
RHCE
NPP3 (Nutanix Platform Professional)
Chuck Parrott
Occasional Visitor

Re: array expand times

We're usually offline for most weekends, so that's when we do all the db maintenance stuff along with backups. We're actually at 99% complete but this is the second extend. Is 2 normal? The first one completed much sooner than expected, it gained ground during low activity periods, but after it finished, it show the drive as unused space. We then started a second extend to incorporate the unused space into the active array.

Another thing we have read about is that we might be experiencing a raid sync problem. Is there any way to tell other than the less than precise method of watching the drive activity lights to see if they blink in unison?

The problem we're trying to solve is that on any given high volume day, we experience times of high disk writing i/o. I'm talking avg disk queue lengths of 50-100 on three spindles. They'll last for about a minute and then subside and be fine (< 1 avg) for another 20-25 minutes. There are no sql jobs running or other abnormal activity on the system that we have been able to pinpoint. Everything I have been able to read points to hardware i/o issue, but the controller reports everything okay. Our first step was to throw another spindle into the array to try and relieve any bottlenecks. We're hoping it will finish today, as it's at 99% on the extend to unused space job.
Steven Clementi
Honored Contributor

Re: array expand times

What service pack are you running on 2000?


I was going to suggest that on the weekend.. you can set the priority to high, but if it is almost finished, then you should be ok.


The process is actually 3 parts...

Extend the Array (you did this when you added the disk)

Extend the Logical Drive (you are doing this now)

Expand the the disk partition (you do this from within Windows using the diskpart.exe utility)

You will need to grab the resource kit (maybe) for Win2k, or find it on their download site. Win2k3+ have it built into the OS build. Shere here for more info...

" http://support.microsoft.com/kb/325590 "


Steven
Steven Clementi
HP Master ASE, Storage and Clustering
MCSE (NT 4.0, W2K, W2K3)
VCP (ESX2, Vi3, vSphere4, vSphere5)
RHCE
NPP3 (Nutanix Platform Professional)
Chuck Parrott
Occasional Visitor

Re: array expand times

Thanks for the info Steven. It did finish and I extended it in W2K (SP4 since you asked) but it actually shows up as 2 distinct drive spaces but 1 volume in disk management. New space is reported correctly. Is that normal? We also had a status message that parity initialization was queued but it hadn't started yet. Not much info on that other than something to trigger I/O on the new drive will cause the parity init to start. I went with a hunch and pulled the new drive out of the array for 30 seconds and put it back in. Now it's started a rebuild process. We'll see when that's done.
Steven Clementi
Honored Contributor

Re: array expand times

Post a screen shot, but it sounds right.

After all... thats what you did.. you expanded the volume, not the disk partition.

The Heirarchy for Smart Array Controllers:

Hardware level:
The Array
--Logical Drives

Software level:
The Logical Drive
--Disk Partitions

3 Distinct Entities.

I.E. You system may have 1 Array with 2 logical drives. The logical drives show up as Disk 0 and Disk 1. Disk 0 is likely your C: drive and Disk 1 is your data drive.

Your data drive probably has 1 partition on it... lets call it D:

You expand the array, than the logical drive... the system still thinks your D: partition is 100MB even though you extended your "Disk1" to 200MB.

Run diskpart to tell the partition to increase to the max size of the new logical drive. (Step 3 as stated above).

As for pulling the drive... you did not need to. As soon as you complete step three... the parity init should kick off and all should be done/well/finite/etc.


Steven
Steven Clementi
HP Master ASE, Storage and Clustering
MCSE (NT 4.0, W2K, W2K3)
VCP (ESX2, Vi3, vSphere4, vSphere5)
RHCE
NPP3 (Nutanix Platform Professional)
Chuck Parrott
Occasional Visitor

Re: array expand times

Here's a capture of our current DM screen. It shows that the new disk is part of the logical volume of D: but shows it as a distinct physical window. Disk size reports correctly. But the display has me concerned that it's acting as clustered and not a contiguous part of the data array.
Chuck Parrott
Occasional Visitor

Re: array expand times

As a follow up, here's a screen capture of the event we're up against. This is a perfmon shot of the disk spikes. The highlighted line is the avg disk queue length. The disk write queue mirrors the the overall disk queue. The disk read is only slightly active, visible in the blue line.

The other things like pages/sec and proc time don't go up. Watching processes during this doesn't red flag any process doing excessive i/o writes. The timing of the event is such that it doesn't match any scheduled SQL jobs. I've tried SQL profiler to see if I could locate any query running at the same time but no standouts.

Frankly this one has us stumped.