User Manual

Features Overview and ConfigurationRev 2.3-1.0.1
Mellanox Technologies
162
Constraints between parameters:
•
dev_loss_tmo, fast_io_fail_tmo, reconnect_delay cannot be all disabled or nega-
tive values.
•
reconnect_delay must be positive number.
•
fast_io_fail_tmo must be smaller than SCSI block device timeout.
•
fast_io_fail_tmo must be smaller than dev_loss_tmo.
3.3.1.1.2 SRP Remote Ports Parameters
Several SRP remote ports parameters are modifiable online on existing connection.
 To modify dev_loss_tmo to 600 seconds:
echo 600 > /sys/class/srp_remote_ports/port-xxx/dev_loss_tmo
 To modify fast_io_fail_tmo to 15 seconds:
echo 15 > /sys/class/srp_remote_ports/port-xxx/fast_io_fail_tmo
 To modify reconnect_delay to 10 seconds:
echo 20 > /sys/class/srp_remote_ports/port-xxx/reconnect_delay
Manually Establishing an SRP Connection
The following steps describe how to manually load an SRP connection between the Initiator and
an SRP
Target. Section , “Automatic Discovery and Connection to Targets”, on page 167
explains how to do this automatically.
• Make sure that the
ib_srp module is loaded, the SRP Initiator is reachable by the SRP
Target, and that an SM is running.
reconnect_delay Time between successive reconnect attempts. Time between successive
reconnect attempts of SRP initiator to a disconnected tar
get until
dev_loss_tmo timer expires (if enabled), after that the SCSI target will
be removed.
fast_io_fail_tmo Number of seconds between the observation of a transport layer error
and failing all I/O. Increasing this timeout allows more tolerance to
transport errors, however
, doing so increases the total failover time in
case of serious transport failure.
Note: fast_io_fail_tmo value must be smaller than the value of
reconnect_delay
.
dev_loss_tmo Maximum number of seconds that the SRP transport should insulate
transport layer errors.
After this time has been exceeded the SCSI target
is removed. Normally it is advised to set this to -1 (disabled) which will
never remove the scsi_host. In deployments where different SRP targets
are connected and disconnected frequently, it may be required to enable
this timeout in order to clean old scsi_hosts representing targets that no
longer exists.