Ever since we upgraded to 3.5U5 from 3.5U4 to supposedly address a buggy Emulex HBA driver causing purple screens, we have been seeing this message in /var/log/vmkernel. It is flooding the logs.
Jan 12 22:26:00 usbo-qvrt-e11 vmkernel: 0:23:16:34.461 cpu6:1047)WARNING: Heap: 1522: Heap_Align(lpfc_740, 65536/65536 bytes, 4 align) failed. caller: 0x90e865 Jan 12 22:26:00 usbo-qvrt-e11 vmkernel: 0:23:16:34.462 cpu6:1047)WARNING: Heap: 1522: Heap_Align(lpfc_740, 65536/65536 bytes, 4 align) failed. caller: 0x90e865 Jan 12 22:26:00 usbo-qvrt-e11 last message repeated 2 times
These are ESX 3.5U5 build 207095 running on HP Proliant DL585 G5s with Emulex LPe11002 HBAs (4 in each). I have 2 ESX servers in this cluster due to our excessive amount of LUNs (153) in the cluster. The other ESX host in this cluster does not see these messages. Due to previous storage issues, the HBA LUN queue depth is increased to 128 and the Disk.UseDeviceReset is set to 0. Due to more issues potentially related to storage, we increased the vmklinuxHeapMaxSizeMB to the max of 40 MB. This was brought up when we had similar message about Heaps but point to vmklinux. I have run HP Firmware v8.6 CD (latest) on there. The HBAs have version 2.82a3 (latest by EMC on the Emulex website). The storage is fiber channel through our EMC Symmetrix. Any thoughts? This all came out to the problematic host purple screening a few weeks ago. Since then, the VMs had all sorts of issues with storage resulting in check disks, blue screens, missing hal.dll and the list goes on.