Disk Enclosures
1748163 Members
3722 Online
108758 Solutions
New Discussion

EVA tuning for use with Falconstor IPStor devices.

 
Mossman
Occasional Contributor

EVA tuning for use with Falconstor IPStor devices.

Hi
I have a customer who is experiencing issues when installing new disks in his EVAs. Each time one is installed it causes the IPStor device to generate a number of 'SCSI command Aborted' errors resulting in the de-stabilising of some attached hosts or even blue screen if under heavy load. The addition policy is set to manual.

IP stor suggest tweaking some settings on the storage although I am not sure how as apart from the disk addition policy and one or two others through Command view, the EVA is self tuning.
As a result, are you able to advise on :
1) What are the considerations relating to blocked I/O during the time new disks are inserted?
2) Are any of the parameters in the attached file tunable or selectable?

i look forward to hearing from you.

Richard Moss
1 REPLY 1
Peter Turek
Frequent Advisor

Re: EVA tuning for use with Falconstor IPStor devices.

I'm an EVA novice but when you add new disks aside from any firmware/software issues you need to worry about at least these two things:

1) When you add a new disk the EVA must "level". It's like changing out a bad disk in a raid array. The redistribution of data has to be done to take up the new space (I think EVA uses 256kb chunks). Leveling can literally take several days during which time your performance will be affected (how much I don't know).

2) When you add disks, you need to take into account how many shelves and how you are arranging the disks into disk groups and for disk protection levels. Ideally you want to add disks in vertical columns across as many shelves as possible.

Depending on how the disk groups are set up that you add new disks to, I'd guess the problem you're seeing may be that the spindles in a disk group are just too busy between host requests and the leveling process.

Do you use perfmon or evaperf to gather stats on how the disk groups and vdisks are performing?

I don't know if the leveling strategy can be tuned.

The scsi command aborted may literally be the response of a host request is too slow and a timeout is reached thus the aborted request. perfmon stats should show you the response times or latency of the controllers, disk group and vdisk, any one of which could be the culprit if it's a speed issue.