FC Path Redundancy Model
Failover Timers
| Parameter | Recommended | Effect |
|---|---|---|
fast_io_fail_tmo |
10 seconds | Time to fail I/O after fabric reports link down |
dev_loss_tmo |
60 seconds | Time before device is removed after persistent failure |
failback |
immediate |
Restore preferred paths as soon as they come back online |
Why these values:
-
fast_io_fail_tmo 10: Fast enough to fail over before application timeouts (typically 30-60 s) while avoiding false positives from transient link bounces -
dev_loss_tmo 60: Allows time for switch or FlashArray failover events to resolve before the device is removed from the kernel -
immediatefailback: Ensures optimized paths are used again after a controller failover resolves - prevents sustained I/O on non-optimized paths
Simulating a Path Failure (Testing)
# Identify path devices
sudo multipath -ll
# Disable one path (replace sdb with actual device)
echo offline | sudo tee /sys/block/sdb/device/state
# Verify multipath reroutes
sudo multipath -ll
# Re-enable the path
echo running | sudo tee /sys/block/sdb/device/state