On 09/01/18 17:44, Parav Pandit wrote:
Rxe driver needs to implement alloc_hw_stats() and get_hw_stats()
functions callback and those counters will be exposed in sysfs at
/sys/class/infiniband/rxeX/ports/<Port>/counters.
They already are exposed, just with wrong content - which leads to the
parsing error. Nevertheless thanks for the hint re. the hw_counters
directory; I missed that and indeed it contains values that are updated
when the rxe device has traffic. So it really is the mapping between
HW device to "IB device" that is missing.
There is a library implemented in golang and used in some of the
orchestration tool as well available at [1]. So you can directly use
the library to integrate in the application if it is in golang.
Now I'm confused even more. :) I really don't want to add special code to
the node collector (especially since I don't really know golang), I was
wondering whether it makes sense to catch the "N/A (no PMA)" values or
do something more sensible like maybe hiding the counters directory instead,
until the hw counters are mapped properly. IMHO no values (yet) would be better
than illegal values. Or instead of exposing wrong (string) values just set
them all to 0, which would "fix" the problem as well.
Once rxe driver implement statistics, those will be available automatically here.
But it already does! See [1], so it's not the collection of values for the
underlying hw device but the connection/mapping to the "IB device" that is
missing, and that was what I was wondering about/searching for.
thanks!
Holger
[1]https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/infiniband/sw/rxe/rxe_verbs.c#n1252
