| Subject: | Better logging when running out of memory |
On a recent support ticket we had a customer who had been having trouble for some time where their DHCP servers stopped handing out leases and logging. The first couple of times that they encountered this they restarted the server without collecting a core dump so we didn't have much to go on for diagnosing the issue.
When they did finally get a debuggable core file to us I was able to determine that the process had gone "non-responsive" because memory allocations were failing. I'm assuming here that out-of-memory condition is why log messages aren't being emitted to let the operator know that the process is out of memory.
I would like to see dhcpd be able to log something when it runs out of memory so that there isn't any guesswork.
When they did finally get a debuggable core file to us I was able to determine that the process had gone "non-responsive" because memory allocations were failing. I'm assuming here that out-of-memory condition is why log messages aren't being emitted to let the operator know that the process is out of memory.
I would like to see dhcpd be able to log something when it runs out of memory so that there isn't any guesswork.