Unsolved
This post is more than 5 years old
1 Message
0
6882
August 8th, 2008 22:00
MD3000i setup linux multipath implementation problem
I have questions about how to setup the MD3000i and correctly load balance access multiple virtual disks. I have two RHEL5 clusters that will be accessing one of two virtual disks on the MD3000i. Virtual disk 0 on LUN 0 will be accessed by cluster 0. Virtual disk 1 on LUN 1 will be accessed by cluster 1. Cluster 0 runs on servers 01-05, cluster 1 on 06-10. On my initial setup I assigned 192.168.1.10-13 to the iSCSI interfaces on the MD3000i. Each of our servers in the cluster has a bonded interface (consisting of two physical connections) which splits the load over two switches. Each server/iscsi-client logs in to 192.168.1.10-13 over the bonded interface. This setup works, but is slow. I also get errors about failed paths in dmesg. I have tried this setup with only one interface (not bonded) and only one switch, with all four iscsi ports on the MD3000i connecting to one switch...I still had the same errors. So I don't think the problem was with how I did the bonding.
I'm just not clear on how to set this up to be fault-tolerant and load balanced. I guess ideally I would want one virtual disk to prefer c0 and the other c1 (so one cluster doesn't affect I/O on the other), but still fail over if it's preferred controller goes down.
Here's what I have in my multipath.conf:
...
devices {
device {
vendor DELL*
product MD3000i
path_grouping_policy multibus
getuid_callout "/sbin/scsi_id -g -u -s /block/%n"
features "1 queue_if_no_path"
path_checker rdac
prio_callout "/sbin/mpath_prio_rdac /dev/%n"
hardware_handler "1 rdac"
failback immediate
user_friendly_names yes
}
}
...
So what is the ideal setup for two virtual disks accessed by two different groups of servers that would give me the highest throughput and still be fault-tolerant?
I realize I'm rambling a bit, so I'll just post this and see what I get. Feel free to ask any questions. I'm new to this, so be gentle!
Thanks.
I'm just not clear on how to set this up to be fault-tolerant and load balanced. I guess ideally I would want one virtual disk to prefer c0 and the other c1 (so one cluster doesn't affect I/O on the other), but still fail over if it's preferred controller goes down.
Here's what I have in my multipath.conf:
...
devices {
device {
vendor DELL*
product MD3000i
path_grouping_policy multibus
getuid_callout "/sbin/scsi_id -g -u -s /block/%n"
features "1 queue_if_no_path"
path_checker rdac
prio_callout "/sbin/mpath_prio_rdac /dev/%n"
hardware_handler "1 rdac"
failback immediate
user_friendly_names yes
}
}
...
So what is the ideal setup for two virtual disks accessed by two different groups of servers that would give me the highest throughput and still be fault-tolerant?
I realize I'm rambling a bit, so I'll just post this and see what I get. Feel free to ask any questions. I'm new to this, so be gentle!
Thanks.
No Events found!



Dev Mgr
4 Operator
•
9.3K Posts
0
August 10th, 2008 17:00
Then from the server side you don't team the nics, but let the MD3000i multipathing software handle the multipathing/load balancing.
You may find some resources about setting up an MD3000i on www.delltechcenter.com.