[evlatests] sequence number errors -- network issues?
Walter Brisken
wbrisken at nrao.edu
Fri Nov 16 11:39:42 EST 2007
I looked through some log data and think I see evidence for sequence
number errors, indicating loss of UDP packet delivery. They seem to come
in groups, as might be expected if there is serious network traffic for a
period. Most of the errors are on tcal docs (as expected -- they are
quite numerous). For example:
54419.idcaf.log:SEQUENCE NUMBER ERROR: jump from 28 to 30, mon point =
idcaf_tcal_seq
54419.idcaf.log:SEQUENCE NUMBER ERROR: jump from 30 to 32, mon point =
idcaf_tcal_seq
54419.idcaf.log:SEQUENCE NUMBER ERROR: jump from 33 to 35, mon point =
idcaf_tcal_seq
54419.idcaf.log:SEQUENCE NUMBER ERROR: jump from 35 to 37, mon point =
idcaf_tcal_seq
54419.idcaf.log:SEQUENCE NUMBER ERROR: jump from 37 to 41, mon point =
idcaf_tcal_seq
These errors all happened within a second of each other
This particular traffic is from mchost to idcaf.
I've seen no evidence of out-of-order sequences, so the problems are all
missing docs, not reordered ones. I'm ignoring for obvious reasons the
sequence number errors associated with test time / clexec. I have not heard
any word from the operators -- are they getting alerts when these happen? My
code doesn't clear this message type automatically.
-Walter
More information about the evlatests
mailing list