On Monday 08 August 2005 8:17 pm, Randall R Schulz wrote:
Hans,
On Monday 08 August 2005 16:13, Hans Ophüls wrote:
...,
it is well-known by sysadmins that there are problems with kernel updates. It seems that the first reboot without power-off may cause sometimes problems. We had Kernel panic after some days, after such a update without power-off (Suse 9.1).
Could I get some references on this "well-known" phenomenon?
...
Regards Hans Ophüls
Randall Schulz
Aside from personal experience from years of hardware/software compatibility testing? Consider that today's wake-on-LAN or -ring functions require power to function. A simple warm reboot will not be the same as pulling the power cord and waiting 30 seconds for capacitors to drain. This does have an effect on many components in a system. Been there, done that many times and have seen the difference. To get a baseline we almost always had to remove all power for x number of seconds or minutes and then power up and re-test. Tons of fun when large SCSI arrays were first being introduced to the PC server arena. Each drive was set to spin up when the controller told it to and that was one-by-one per bus and up to 4 busses per controller. Those first 1GB SCSI drives were HUGE and still too small. Can you say single-ended versus differential and passive versus active termination? How long was that cable supposed to be to be in spec? At most we'd mention removing power in the manuals to "zero" out any stored charge(s) that should have been drained. There was no way to effectively remove all power through mechanical switches since there was a need to monitor power throughout the system if it was plugged in but not turned on. Had to know if there was AC to the system in a lights-out scenario due to spinning up 28-60 SCSI devices/controller in a controlled manner. Stan