Dell Emulex Family of Adapters - Manual (Page 966)

Emulex Drivers Version 10.2 for Linux User Manual P010081-01A Rev. A
Appendix A. iSCSI Driver
966
To permanently set large_io on every iSCSI boot, edit the /etc/modprobe.conf file, type
options be2iscsi eto=0 ldto=0 large_io=512 im_policy=1
After adding the options line, save the file, and then rebuild initrd.
iSCSI Error Handling
Error Handling using ETO and LDTO Parameters
The goal of iSCSI error handling is to be tolerant of link-level and target-level failures
up to configured timeout values so that I/O errors are not seen by the application or
operating system.
The error handling is triggered under the following conditions:

Loss of connection to the target due to target and/or network disconnection at
the target.
If the driver has I/O requests pending with the target and the target becomes
unavailable (due to the target going down or failing over, or network issues at
the target), the driver queues up the I/O requests internally for a configured
period of time. The threshold value of this period is the ETO value.

Loss of immediate link to the initiator (such as cable disconnect or port failure)
The adapter firmware detects and notifies the driver of a loss of the link. When
this happens, the driver queues up the I/O requests internally to a configured
period of time so that the operating system does not see I/O errors. The
threshold value of this period is the LDTO value.
When the configured ETO or LDTO value is reached, and the initiator is still unable to
connect to the target, the driver fails all I/O requests. At this point, I/O errors are seen
by the application and operating system.
Note: Following a link up, switch ports can take a long time to initialize and go to the
forwarding state. Because of this, additional time should be added to the ETO
and LDTO settings to eliminate I/O disruption or target unavailability. If the
switch port is connected to a single host, then the PortFast mode can be enabled
on the switch port to eliminate delays when it transitions to the forwarding
state.
Error Handling Under MultiPath I/O (MPIO) and Cluster
Configurations
In an MPIO or cluster configuration, fault tolerant software is present on the system
that makes the iSCSI driver error handling redundant. These configurations also
require that I/O errors be reported as soon as they are detected so that the software can
fail over to an alternate path or an alternative node as quickly as possible.
When the iSCSI driver runs under these configurations, the error handling
implemented in the driver must be turned off by setting the default values of LDTO
and ETO to zero. The changes take effect during the next driver load.