I have a pecular problem. On my ME, if usage slowly builds up over hours, no problem. However, if it is suddenly hit in less than a few minutes, it kills the ME.
I have determined that the sockets go wonky, some keep working, others may start about 20% of the time, the rest of the time I get error messages like “no socket available” or “no response from remote peer”. I have 50 sockets allocated for the class. In another class I have 60. Then a couple using 2-3 sockets. I have heard that there are 128 available from some sources, 192 from others. In any case I am using about 120.
You may believe you are using 120, but by default there are 7 - 8 preallocated by the OS (DHCP, ADDP, serial ports …) If you have allocated 120 that yo know of then yo are probably over the 128 limit.
I figured out it wasn’t the sockets, it was the program size (1.6M). I killed six 9210s before I realized that they were only partially updating when I FTPed in image.bin and now I get the constant dual LED on boot.
I have reduced the footprint and have a working one now, but the only working ME I have left is the JTAG ME, simply because I can override it. sigh Now I have to dig up the recovery procedure, desoldier those MEs, and fix 'em.
You would think that the image update would be more graceful on failure.