I am diagnosing several systems where tx_restart_queue continually increases,
although overall system load and network throughput isn't very high. I was
hoping that someone could confirm my understanding of what this counter
actually means.
>From what I can gather tx_restart_queue represents the number of times that
>transmits were delayed because the ring buffer is full. No data is dropped or
>lost, it is only delayed until there is room in the buffer, i.e. after the
>next interrupt clears the buffer. The actual worst case delay would be
>roughly the interval between interrupts. Is this correct? I mainly want to
>be sure that we are not losing data... no other problem counters (such as
>rx_no_buffer_count or rx_missed) are incrementing on these systems.
I did see an earlier thread where it was recommended to tune the interrupt rate
(rx-usecs) and possibly tx descriptors in order to address this issue. So far
I'm having some success in doing this. I am seeing this on several systems
using either 82571 (e1000/e1000e) or 82576 (igb) parts on RHEL4/5.
Thanks,
Mark
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
