Lars NoodC)n wrote: > bofh wrote: >> I think Marco's point was that if there are crashes, lockups, etc, it >> is a pain in the ass not to have console access, or to be able to >> unplug the power and reboot into a working config/kernel, etc etc. >> ... > > Access to a second box, for control, which has both serial and Ethernet > connections to the development units is the first step. > > What else is there on a wish-list for being able to do kernel-level work > remotely? > > What architecture-specific troubles are common? I notice some of the > PPC and AMD64 units I have do not have serial output. > > As far as the power-cycling of hardware goes, I'm looking into GPIO on > the motherboard to control optocouplers (or equiv) to control power > relays which can then switch the wall current for some development units > on or off. For equipment with hardware reset, the relay can be skipped. > These would be controlled via the control box's GPIO using gpioctl. > > I ask because I'm trying to set up an environment where kernel-level > work can be done remotely. Having more perspectives will help. > > regards, > -Lars > >
Lars, coupla things that might be useful, first off it bears mention that quite a few higher-end machines (eg Sun Fire T1000) have a lights-out management device or card of some sort which allows pretty much full control over the machine over either ethernet or serial, or both. In the case of the T1000, its only interfaces are 4 ethernet jacks for the machine, and 2 RJ45's, one serial and one ethernet for the ALOM. A machine of this type wouldn't need a management machine at all, everything can be done over ethernet. Given it doesn't have peripheral or monitor outputs, there is in fact no difference between being there and being on another continent. The T1000 is the extreme case, but both HP and Dell to name two companies provide management cards which give a certain amount of remote control. Googling for Dell DRAC and HP iLOM may be useful to you. The other thing is that the guys over at CoreBoot have spent some time addressing low-level remote control of motherboards for testing their BIOS code. I don't know if they still do, but they used to have an automated system which built, flashed and booted their code on a variety of motherboards. I think Google provided some of the resources. I don't know if its the sort of thing that you're looking for, but their work might give you a head start on being able to reset systems with GPIOs. si1entDave
