Unsolved
This post is more than 5 years old
2 Posts
0
90324
July 10th, 2013 15:00
Multiple Servers accessing a single iSCSI target
I would prefer a true Windows Failover Cluster, but our customer has some unique needs. I've made this work with SAS attached storage in the past, however this partctular project specified that iSCSI be used. My experience with iSCSI is limited to traditional deployments in which only a single server will connect to an iSCSI target.
The "cold failover" method would utilize two servers and a SAS attached storage device, all with redundant cabling. Both servers would have idendical images and the storage array would house the SQL database and application files. The "backup" server would remain powered down until it was needed, at which point they would simply power the bad server down and bring the spare online.
On the current project, I have deployed a pair of Poweredge 720s and a Powervault 3200i that is using direct iSCSI connetions (MPIO). I was able to intinailize the disk and format my 3.5TB partition (GPT) and write data, however when I killed power to the "primary" and brought the "backup" online the data was nowhere to be found.
I know it's not exactly best practice to have multiple hosts connecting to to the same NTFS partition, but since only one is conneted at a time it has worked with SAS connections. Is this an iSCSI limitation keeping this from working as intended, or a limitation of the 3200i?
DELL-Sam L
Moderator
•
7.6K Posts
0
July 10th, 2013 16:00
Hello V1Corp,
Your setup will not work without having cluster management in the middle. It is a limitation of iSCSI. Unlike SAS when you have multiple servers accessing iSCSI storage & no cluster in the middle then you get corruption. It is hard to say how long before you will get corruption. You didn’t state in your post but when you brought back online server A were you able to access the data without getting any errors?
Please let us know if you have any other questions.
V1Corp
2 Posts
0
July 12th, 2013 12:00
Yes, when server A was brought back online, I was able to access the data without corruption. The issue that I'm still having in our testing environment, is that when server A is taken offline and server B is taken off the shelf and racked in it's place the data doesn't appear.
I currently have two controller ports on the md3200i allocated for each server, and all four nics are in separate subnets. In the event of a total server failure, what is the preferred method for reconnecting to the existing data when swapping in a total replacement server?