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
If we use the BatchExportSpanProcessor combined with the
JaegerSpanExporter and use instrumentations that add a lot of metadata
to the spans like sqlalchemy, then we run occationally into the "Data
exceeds the max UDP packet size" warning causing dropped spans and
incomplete data.
The option to reduce the general batch-size to a very small number (in
my case >30) may cause a performance issue as the worker thread of the
batch exporter gets very busy. Instead this change allows the user to
ask the exporter to split oversized batches when they get detected and
send the splits separately instead of dropping them. Depending on the
usecase this is a better option than reducing the batch-size to a very
small value because every now and then they contain a couple of large
spans.
0 commit comments