Duet Memory on NX Controllers

[Deleted User]

Several recent projects leveraging duet modules on NX controllers have found java router lockups after varying amounts of runtime. A simple first step in remediation is to up your duet memory to 128mb (default was 64mb). For sites we were tracking, this one change resolved the majority of issues. It appears that some sites would run right up against the 64mb limit and then if it crossed the line, you were toast. Moving the line by increasing the allocated memory seems to have helped in those surge instances and avoid catastrophic lockups.

Certainly not a silver bullet for all processor lock-ups, but something to try if you find yourself in a frustrating situation you cannot explain with a NX + duet combo.

STEPS TO CHANGE
telnet or ssh into controller
superuser
set duet memory
128
reboot

emdx71

would it be possible to do this in code via a send command and with setting a flag when its been done or is this something that has to be done via telnet/ssh?

[Deleted User]

The command is only available via telnet/ssh, but there is nothing stopping you from writing a function that fires at boot to telnet into yourself, get the memory, then react to the result.

John Nagy

@emdx71 said:
would it be possible to do this in code via a send command and with setting a flag when its been done or is this something that has to be done via telnet/ssh?

Note that once done, it stays at the new value permanently (surviving boots and new code) until a new setting is made.

Marc Scheibein

@HARMAN_Chris said:
The command is only available via telnet/ssh, but there is nothing stopping you from writing a function that fires at boot to telnet into yourself, get the memory, then react to the result.

There are functions to to get and set Duet Mem size... DUET_MEM_SIZE_GET to request, and DUET_MEM_SIZE_SET(). But the SET is only documented for up to 64M

SCamphaug

I've absolutely noted this issue in rms deployed systems lately. 96m seems to hold on the mcp keypads and my enovas seem to be ok running at 128. It's fairly easy to do in mass via Python