I'd welcome some improvements to kdump/netconsole:
- Could crash somehow guesstimate from a running system how much disk space will be needed?
- Would be nice to be able to use swap space as dump device to avoid needing to have empty space "in reserve" just in case you panic some fine year. Especially on VMs, this adds up to a lot of wasted space (without thin provisioning which isn't always a possibility).
- Half the time all I need to know is the EIP - couldn't the panic output be logged someplace persistent without needing netconsole and a remote log server?
- Finding the debuginfo packages is sometimes a challenge (try to find kernel-debuginfo-common-2.6.128-1.14.el5 through RHN, I dare you). Couldn't they just be in a normal yum channel?