For quite some time I had been experiencing power-off problems on a HP Proliant server running Solaris (or OpenSolaris). Most of the time, the
poweroff or init 5 commands will not cut the power and the machine will hang and stay up with fans spinning at full speed. The only solution was to manually power cycle or use LOM to shut the machine down. This has caused data corruption problems for me several times, specially when the HP SmartArray batteries get discharged.
Turns out there is a HP support document, SHUTDOWN PROCEDURE REQUIRED for ProLiant Server Running Sun Solaris 10 to Properly Flush Cache on Smart Array Controller Prior to Shutdown that precisely describes my situation. It seems it’s an interaction problem between Solaris and the
cpqary3 drivers (2.1.0 or older) that causes buffers in the HP SmartArray controller to not be flushed on shut down. If the batteries drain, eventually data loss might occur. And precisely, since I was using
cpqary3 driver version 2.1.0, I have just upgraded to
cpqary3-2.2.0 to see if this solves my problems for once and all. We’ll see 🙂