You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In the recent times, I see that UDP packets are being dropped in kernel due to overlimits. I was reading a bit to understand the term "overlimits" in tc qdisc. From the study I infer that, it is the delay caused by qdisc and that's the reason packets are dropped when the txtime expires waiting in queue to be dequeued. Am I right? Or correct me if am wrong?
May I know why this scenario comes in?
The text was updated successfully, but these errors were encountered:
During development we had a DROP_IF_LATE cmsg flag that we could set per packet to select if that packet was allowed to be dequeued even if late ("expired"). But after a very long discussion with the upstream maintainers, the agreement was to remove that flag and keep this behavior.
From my experience, an increasing "overlimits" usually mean some problem in time synchronization, or that for traffic with lots of "burstiness" that the network card cannot dequeue packets in time.
I have ensured that time synchronization does not exceed few ns accuracy while the test is being performed. At times, packets do not get dropped or marked as overlimits even when the udp_tai application is running at 100us cycle interval. So I am not able to conclude if burstiness could be the only reason.
Hi @vcgomes,
In the recent times, I see that UDP packets are being dropped in kernel due to overlimits. I was reading a bit to understand the term "overlimits" in tc qdisc. From the study I infer that, it is the delay caused by qdisc and that's the reason packets are dropped when the txtime expires waiting in queue to be dequeued. Am I right? Or correct me if am wrong?
May I know why this scenario comes in?
The text was updated successfully, but these errors were encountered: