- Posted In
- Red Hat Enterprise Linux
I know, judging by several threads I read on the Fedora mailing lists, this question likely invokes religion, but, since I'm in the process of helping our build team come up with standards for our soon to go to production RHEL 6 deployments, I'm feel compelled to ask, any way.
Prior to my current job, much of my UNIX work was done on Solaris based systems. As such, I was used to the /tmp mountpoint being a tmpfs filesystem rather than a "real" filesystem. Prior to the forthcoming RHEL 6 deployment, my current employer's use of RHEL was complimentary to Solaris - it's use was a beneficial side-effect of trying to push everything onto ESX and Oracle's bone-headed decision to effectively discontinue Solaris x86.
The RHEL 5-based standard build that was put together by prior engineers used /tmp as a "real" filesystem (ext3 on an appropriately sized LVM volume). I know that Linux has had the capability to support /tmp on tmpfs for quite some time. I know that some distros have even flirted with making that their standard (Fedora 18 being one of them, judging by the gnashing of teeth on one of the relevant discussion groups). So, what I'm looking for are two things:
- Why would(n't) I want our RHEL 6 builds to switch to the more Solaris-y /tmp on tmpfs - especially in ESX-hosted environments where disk-contention is likely to be of greater concern than memory-shortage
- How would I convert my KickStarts to set up /tmp as tmpfs rather than as a real filesystem? Can it be done in the main sequence-block of the KickStart configuration or would I need to do it as a %post operation (and any gotchas/ordering-issues in doing so)?