Friday, May 30, 2008

Technically Accurate, But Not Politically Astute

The support team for a security appliance I use recently responded to a periodic lockup issue by informing me that "When the appliance starts to overload, random processes are killed to free up memory". A co-worker noted that this was "technically accurate, but not politically astute" - he's entirely right. While we would be upset if the vendor didn't provide a reasonable answer, this one causes worries. Why?

In part, because the answer is somewhat technically accurate for a Linux based appliance, and it may even be related to the actual issue. It isn't, however, something that a customer wants to hear, particularly for a hopefully bulletproof, nominally trusted security appliance. One imagines that the vendor has done various clever things, like protecting their critical processes, and has taken steps to ensure that if they do die, that they come back automatically. In this case, a partial answer wasn't reassuring.

I didn't expect an in depth lecture on oom_kill.c, but I will expect a follow up with a good answer as to why the device can hard lock without the central control system notifying me.

No comments: