One of our clusters we tried out used C2100s and these LSI 9260-8i controllers. They ran fantastically out of the gate, but we started to run into some issues. I can attest to the fact that our rack of servers ran hot. In pulling an over nighter it would get cold in the facility and I’d stand behind that rack because the heat coming out of over a dozen servers with a dozen disks running and processors running full bore kept my fingers from cramping up. This wasn’t heat senor warning hot, but it definitely put out some warmth. Later on when we started to have issues we’d pull the controller out and find the heat sink sitting on top of the card not attached in any way. What I think the problem was is that these controllers have the heat sink held on by plastic clips and springs and as our servers would run warm, not to the sensor heat warning level, eventually the clips would melt leaving the heatsink sitting there on top of the controller.
2 replies on “LSI 9260-8i fails”
I’ve had a few of the similar IBM version of this, the M5015, where the heatsink has fallen off as well.
I wonder how common the problem is.
Yup, it’s just happened to me as well. The plastic push-screws had become brittle to the point where’d they’d broken in multiple places within each and the heat-sink was just found sitting loosely on top of the board. I was getting intermittent loss of storage etc. The silver lining is that it’s prompted me to migrate away from EXT3 via the HBA to ZFS on the native 6Gbps ports.
John