Agent Memory Usage

How does Agent message memory usage compare to other server API calls?

I have a firmware running with about 8k imp.getmemoryfree() but if I trigger an event which sends a few quick Agent.send() calls (on the order of 5, each with a table as the second argument) I can easily chew threw that and make the imp crash (with no out of memory error printed to the log). Just curious if they are heavier than an inputport/outputport or server.log messages in the underlying OS TCP buffers.

If there’s no out-of-memory error in the logs, maybe you’ve found a different problem. I’ll look into it, thanks for reporting!


Just to add a little bit more information - I see the imp crash without any reason listed in the log happen quite often. Once the imp is booted I typically run at about 8000 bytes reported by imp.getmemoryfree(), but during the boot process I’ve seen it log less than 2000 (and have had to do some hacky things to get the imp past the boot up).

Another thing I’ve noticed that happens regularly is that if I change models and click run the imp will fall download new code, configure, then fall offline, followed by a red LED blink/crash. The imp then reboots and will either enter an endless reboot/crash cycle or will run the new code and be fine - I haven’t pinpointed exactly what causes it either way.

I can start a support ticket and provide you my code if it is easier to debug that way.