Long CPs dominated by P2_FLUSH during SnapMirror Finalizing phase
Applies to
ONTAP 9
Issue
Long CPs occur on the SnapMirror(XDP-type relationship only) destination at the Finalizing phase, Long CPs are dominated by an elongated P2_FLUSH phase without any identifiable Disk or CPU bottlenecks from Perfstats or PerfArchives.
Below are some of the possible symptoms:
- Long CP > 30s
- Back to Back CPs along with high write latency
- CP dominated by P2_FLUSH phase
- CPU util & Disk util is NOT high
- SnapMirror XDP transfer is in the Finalizing phase, no matter it's a baseline transfer or an incremental transfer