oVirt环境下Fibre Channel直连MPIO配置咨询

Hi Matthias, let's break down your questions step by step since you're new to this space—no worries at all, this stuff can feel overwhelming at first!

Do you need to modify multipath.conf? #

Yes, in most cases you’ll want to tweak this file, especially for Dell EMC Fibre Channel (FC) storage. The default multipath.conf included with CentOS 7 has basic rules, but it may not have optimized settings for your specific Dell EMC array (like Unity, PowerVault, or PowerMax). Without adding a device-specific section, multipathd might not correctly group your FC paths into a single multipath device, or might use suboptimal failover/load-balancing behavior.

Will CentOS automatically recognize the two paths? #

It might, but it’s not reliable without proper configuration. By default, multipathd scans for block devices that share the same LUN ID, but if your Dell EMC array’s vendor/product strings aren’t matched in multipath.conf, it might treat each FC path as a separate device instead of merging them. To confirm, you can run multipath -ll before modifying the config—if you don’t see a merged multipath device listed, you’ll need to add the device rules.

Active/Passive vs. Active/Active with Load Balancing? #

This depends on two things: your Dell EMC storage array’s configured mode, and the parameters you set in multipath.conf.

• Most modern Dell EMC FC arrays (like Unity or PowerMax) support active/active mode out of the box, where all paths are usable simultaneously. In this case, you can configure multipath to use a load-balancing policy like round-robin (rr) to distribute I/O across both paths.
• If your array is set to active/passive, only one path will be active at a time, with the other on standby for failover.

To set this up, define the policy in your device section of multipath.conf—for example, path_grouping_policy multibus combined with rr_weight uniform enables round-robin load balancing across all active paths.

oVirt-Specific Notes #

While oVirt’s docs focus more on iSCSI, the FC multipath workflow is straightforward once the host-level config is correct:

• Ensure multipathd is enabled and running on all 4 oVirt hosts before adding them to the cluster or configuring storage domains.
• When adding a storage domain in oVirt, select the merged multipath device (located under /dev/mapper/[device-name]) instead of individual /dev/sdX FC paths. oVirt will recognize the multipath device and leverage its failover/load-balancing capabilities automatically.
• Never configure storage domains on individual FC paths—this will break high availability and failover.

Quick Configuration Steps #

1. First, identify your Dell EMC array’s vendor and product strings using:

   lsscsi
   # Or for a specific device:
   udevadm info --query=all --name=/dev/sdX | grep -E "ID_VENDOR|ID_MODEL"
   

2. Edit /etc/multipath.conf and add a device section tailored to your array (example for Dell EMC Unity):

   defaults {
       user_friendly_names yes
       find_multipaths yes
   }
   devices {
       device {
           vendor "DellEMC"
           product "Unity"
           path_checker tur
           path_grouping_policy multibus
           failback immediate
           rr_weight uniform
           no_path_retry queue
       }
   }
   

3. Restart the multipath service:

   systemctl restart multipathd
   systemctl enable multipathd
   

4. Verify the merged multipath device exists:

   multipath -ll
   

If you run into issues, double-check your Dell EMC array’s documentation for the exact multipath parameters recommended for your model—they can vary slightly between arrays.

内容的提问来源于stack exchange，提问作者MatthiasAUT