[etherlab-dev] Possible problem in 1.5.2 after 30+hours test (not solved yet)

[Johnny Rasmussen]

Hi Etherlab-dev.

Just wanted to let you about a problem I'm currently chasing.

We have 2 domains:
EtherCAT 0: Domain0: Logical address 0x00000000, 373 byte, expected working counter 18.
EtherCAT 0:   Datagram domain0-0-main: Logical offset 0x00000000, 373 byte, type LRW.
EtherCAT 0: Domain1: Logical address 0x00000175, 548 byte, expected working counter 3.
EtherCAT 0:   Datagram domain1-373-main: Logical offset 0x00000175, 548 byte, type LRW.

Domain0 is updated cyclic every 10ms to 6 slaves
Domain1 is updated cyclic every 4ms to 1 slave
On domain 1 there is a single SDO download to the slave every 1 second. (change of a simulator value)
There is no EoE running.

This runs without error for ~40 hours but then I get a 'EtherCAT DEBUG 0: UNMATCHED datagram:' with a dump of the EtherCAT telegram and all SDO exchanging stops working.
When debugging with 'ethercat debug 1' I can see the SDO exchange has finished correctly when the UNMATCHED error happens.

It seems that the master stops checking for new SDO requests after this error. (SDO request stays in 'busy/queued' forever)
Also, when I exit my C application and the master returns to the EtherCAT-IDLE thread there is no data flowing when you monitor with 'ethercat ma' (Rx/Tx stays the same) so it seems locked.

I have to restart the master to get it up and running again.
The UNMATCHED error will repeat it self every 30-40 hours if I keep the system running.

If anyone has experienced anything like this I would be happy to know it :)
Best regards,
Johnny Rasmussen
DEIF Wind Power Technology
Software Developer
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.etherlab.org/pipermail/etherlab-dev/attachments/20130823/7b7f3539/attachment.htm>