Critical Problem for NETAPP users with 6.4 release of device-mapper-multipath
Updating device-mapper-multipath to 6.4 w/out updating kernel on your NETAPP system and it either won't boot or all your multipath devices break
The problem affects all users of NETAPP direct-attached (not network-attached) storage, i.e. iSCSI and FC but not NFS/CIFS etc.
The mpath target feature "retain_attached_hardware_handler" was added to the kernel and userspace multipath-tools for RHEL-6.4:
The kernel change was merged in 2.6.32-299.el6
The updated package will attempt to set this feature on NETAPP devices (the only built-in controller config to use this) but does not check the target or kernel version. This causes an error during table load and prevents multipath maps from being created on older kernels:
multipath: Unrecognised multipath feature request ioctl: error adding target to table dracut Warning: LVM vg/lv not found
How to Reproduce the Issue
- Configure a multipath device to use retain_attached_hardware_handler
- Boot a kernel < 299.el6
- 3. Attempt to configure multipath maps
Workarounds
There are three ways to work around this:
- Update to a 6.4 kernel that supports this device-mapper feature i.e. latest available 2.6.32-358.2.1.el6
- Downgrade the device-mapper-multipath package if the update to a RHEL-6.4 kernel is not possible
- Specify a custom NETAPP configuration in /etc/multipath.conf that removes the retain_attached_hardware_handler feature for these devices, e.g.:
devices { device { vendor "NETAPP" product "LUN.*" path_grouping_policy group_by_prio getuid_callout "/lib/udev/scsi_id --whitelisted --device=/dev/%n" path_selector "round-robin 0" path_checker tur features "3 queue_if_no_path pg_init_retries 50" hardware_handler "0" prio ontap failback immediate rr_weight uniform rr_min_io 128 rr_min_io_rq 1 flush_on_last_del yes fast_io_fail_tmo 5 dev_loss_tmo infinity retain_attached_hw_handler no detect_prio yes } }