Any thoughts on what the answer to this question
should be (and why)?
It has been determined that your server has more than
4GB of memory.
Due to a kernel performance issue on some 2.4
kernels, closing CCISS
device nodes may take lots of CPU time on servers
that have a large
amount of memory. By answering 'y' to the following
question, agents
will keep CCISS device nodes open without ever
closing them to achieve
better performance. Please be advised that by doing
so, you may have
to stop agents before you can configure the disks
attached to CCISS
controllers with such tools as fdisk or cpqacuxe.
Do you wish to keep CCISS device handle open (y/n):
I would recommend answering "YES" to this question. This is a workaround implemented in the storage agents for a potential performance issue, which is seen only on servers having more than 4G of memory running 2.4 kernels.
The storage agents usually has a polling cycle of 15 secs, where it repeatedly issues SCSI IOCTL open/closes during the Storage Agents data collection interval. This in turn consumes a lot of CPU, which might even result in COS hangs and in turn generating ASRs. Answering yes to the above question will keep the device nodes open, and works around this potential issue. And again, this issue is seen only on servers which have more than 4 GB of memory.
If you have already installed the agents by answering "N" to the above question, follow these steps to implement the workaround --
1. Open the "/opt/compaq/cma.conf" file for editing on the ESX Service Console
2. Locate the variable "cmaCloseCcissHandle". This variable is set to "ON" by default
3. Change the value of this variable to "OFF"
4. Restart the agents using the command, "service hpasm restart"
