[ofa-general] problems with epoll and ipoib (memory leak and cpu creep)

[Roland Dreier]

 > When we run the application using poll, things work fine (except for
 > the occasional dropping of packets due to the large set of fds passed
 > to poll).  To try to remedy the problem, we switched to epoll.  Now,
 > as the program runs, the cpu utilization rises over time towards 100%
 > and the memory usage grows as well.

I think the first thing I would check is that your program is using
epoll correctly.  It is really hard to see a way that ethernet vs. ipoib
could affect epoll -- with both ethernet and ipoib, the IP packets go
through the whole network/socket stack before getting to any poll or
epoll handling (and both poll and epoll use the same file method
anyway).

Do you have any way to find out where the growth in memory use is coming
from?  Do you mean that your application is allocating more and more
memory, or is it more like a kernel memory leak?

 - R.