I'm putting this in the beta forum because it's related to the beta testing we're doing. It MAY not be related to the beta code itself, though.
In doing some throughput testing, I discovered I could reliably reset one of my nodes by SCP'ing a file bigger than about 8 MB to it. It doesn't matter what kind of file it is. I verified there was enough space in /tmp and SCP'ed the file there. Smaller file sizes don't reset it. The file transfer gets about 98% done, stalls, then the node resets. Running top on the node in question during the file transfer doesn't show anything remarkable. load is low and it's not running out of RAM.
The reset occurs regardless of whether I do the file transfer via a wireless link or across the DtD link.
The node in question is a Bullet M2. It's running the latest beta code and appears to be functioning properly.
I can transfer the 8MB file in question through that node to all my other beta nodes with no issues (Rocket M2, and two TP-Links CPE210s).
Any pointers on how to troubleshoot this issue?
Thanks.
In doing some throughput testing, I discovered I could reliably reset one of my nodes by SCP'ing a file bigger than about 8 MB to it. It doesn't matter what kind of file it is. I verified there was enough space in /tmp and SCP'ed the file there. Smaller file sizes don't reset it. The file transfer gets about 98% done, stalls, then the node resets. Running top on the node in question during the file transfer doesn't show anything remarkable. load is low and it's not running out of RAM.
The reset occurs regardless of whether I do the file transfer via a wireless link or across the DtD link.
The node in question is a Bullet M2. It's running the latest beta code and appears to be functioning properly.
I can transfer the 8MB file in question through that node to all my other beta nodes with no issues (Rocket M2, and two TP-Links CPE210s).
Any pointers on how to troubleshoot this issue?
Thanks.
Transferring the file through the node is significantly different than transferring the file onto the node.
But to rule out any issues a serial console would be needed likely to see what is going on. I can reproduce on a nanostation that doesn't have a console so I can't get the logs out of it at the moment to be sure that is what is happening, but that is my best guess considering I saw the process get killed (which would be the system trying to control the process) and then a couple moments later the system rebooted (watchdog must of timed out I'm guessing because of lack of resources)
Orv