RCWRCW IT TrainingFree hands-on labs & simulators← Back to home
Pure Storage · Troubleshooting guide

Pure Storage FlashArray Troubleshooting: Capacity Alerts, Hosts Missing Volumes and Replication Lag

Quick, correct responses to the FlashArray alerts that page storage admins: effective capacity pressure, host connectivity, and async replication health.

Published August 28, 2026 · RCW IT Training

Capacity and volumes

Array alerts as capacity crosses 80 or 90 percent

Look beyond raw numbers: in the Capacity view split volumes versus snapshots versus destroyed-but-not-eradicated data (the 24-hour eradication timer). Old snapshot policies are the usual hoarder - tighten or expire them. Plan eradication of destroyed volumes and compare the alert threshold against your real growth rate.

Host cannot see a newly connected volume

Confirm the volume is connected to the correct host or host group and that the host IQN or WWPNs match the array's view. Then rescan on the host - iscsiadm -m session -R for iSCSI, a fabric rescan for FC, reconnect for NVMe-oF. Multipathing should show all expected paths; if not, compare the host initiator list with the array's zoning and portal groups.

Replication and upgrades

Async replication lag growing or snapshot transfers failing

Check replication link health and bandwidth in the array UI; one huge volume change can exceed the window. Verify the network path - MTU and firewalls on replication ports - and that the target is not saturated with its own I/O. If a policy repeatedly misses RPO, split the protection group or add bandwidth, and alert on lag, not just hard failure.

After a Purity upgrade, one host reports timeouts

Review the release notes for HBA and driver minimums - several historical issues were host-driver related, not array related. Collect the support bundle, compare pre- and post-upgrade path states, and open the case with the array ID; FlashArray support moves fast when given the bundle.

Key takeaway: On FlashArray, snapshots and the eradication timer explain most capacity surprises; host visibility is a connect-rescan-multipath walk; replication needs lag alerting, not just failure alerting.