Enterprise Storage · Troubleshooting guide

Enterprise Storage Troubleshooting: Latency Spikes, Missing LUNs and Multipath Path Failures

A SAN admin's diagnosis order for slow datastores and LUNs, hosts that cannot see new volumes, and degraded multipathing.

Published August 28, 2026 · RCW IT Training

Latency and performance

Storage latency spikes at predictable times

Correlate with scheduled work: backups, snapshot creation, replication windows, and array-side scrubs or RAID patrols. Compare queue depth and latency per LUN on the array with iostat -xz 1 or esxtop on the host. If one host suffers while others are fine, the problem is host-side - HBA, driver, or multipath policy - not the array.

All hosts slow after a controller failover event

Verify the array returned to ALUA optimized paths: with one controller owning everything, traffic crosses the interconnect and latency doubles. multipath -ll should show active/optimized on preferred paths; rebalance or fail back per the vendor runbook, and investigate why the failover happened - power, cabling, or firmware.

Visibility and paths

New LUN mapped but the host does not see it

Walk the chain: zoning shows the host WWPN and target port, the masking or host group includes the host, then rescan - echo "- - -" > /sys/class/scsi_host/host*/scan on Linux or a storage rescan on ESXi. multipath -ll must then show the device; if paths are missing, check each switch fabric for the target WWPN and the HBA link state.

multipath shows paths failed or faulty

Identify the layer: dmesg for SCSI sense codes, HBA firmware logs, and switch port error counters (CRC). Re-seat SFPs and swap cables first - they are the classic causes - and keep checker and prio settings aligned with vendor recommendations, for example service-time 0 with ALUA.

Key takeaway: Split SAN latency into array-side versus host-side by comparing one host against all hosts, and treat a missing LUN as a zoning, masking, rescan, multipath checklist.