Closed jaycedowell closed 7 months ago
This seems to be related to flow control on the switch at some level. If I set "flowcontrol send off" for ports 31 and 32 things seem to be better. However, there also seems to be continuous low-level packet loss on the T-engine now.
I thought packet pacing might help with some of the T-engine packet loss but it doesn't really seem to do anything.
What did seem to change things was moving half the T-engines from verbs to "standard" packet capture. No idea why that matters. Maybe some kind of resource limit in verbs or on the cards? It's still not perfect but it's better.
Since this problem seems to be resolved I'm closing this and moving the T-engine stuff to #30.
This seems to be a feature of running everything on a single server. It's not clear what is going wrong but it doesn't seem to be related to the TBF buffer. Could be a (another) problem with
RetransmitOp
.