[erlang-questions] VM ends with 'killed'

[Robert Raschke]

Hi Gordon,

On Mon, Mar 14, 2011 at 8:31 AM, Gordon Guthrie <gordon@REDACTED>wrote:

> Folks
>
> I have been trying to resolve some run time issues. We are running a
> big data load and the VM was crashing due to memory running out.
>
> The first issue was a runaway message queue - so we put in a limiter
> that checked that processes message queue and waited til it cleared if
> there was overload.
>
> The problem moved to another process whose heap would grow and grow.
> Adding a check on the heap size and forced garbage collect appeared to
> be the answer.
>
> But now when I run the load, after adding most of the data the VM
> suddenly ends with the stark message:
> Killed
> [End]
>
> There is no crash dump. What do I do now? Where do I start debugging this?
>
> Gordon
>
>
This'll be on Linux, yes? I've seen this happen when Linux decides that any
one of the processes just needs to die, in order for the system as a whole
to continue. If I remember correctly it has to do with running out of
physical memory, and Linux as a last resort picks victim to kill. And it
kills it outright.

There may be heuristics as to which process it chooses.

Robby