Open tisba opened 6 years ago
@nniclausse do you have any idea how to track this issue down? I haven't figured out a good way to reliably reproduce this yet unfortunately :/
Ping @nniclausse. Any idea what is going on here?
One possibility is to receive the http status just before the stats are dumped to the disk; the end of the request car occur a few msecs after the status is known, and therefore logged 10 sec later
Can/should we change this behaviour? Like counting the request only of the the status is known (or we know that we cannot determine the status)?
I recently observed that we have snapshots in our tsung.log that have 0 reported requests for this 10 sec segment, but reported http status codes.
A slightly simplified example snapshot looks like this:
This was produced by tsung build from 68e509fedc.
Status codes are reported to
ts_mon
here https://github.com/processone/tsung/blob/68e509fedc703a7a2a444c90a8f2df882d517d8a/src/tsung/ts_http_common.erl#L615 in ts_http_common:parse_status/2 andrequest
here https://github.com/processone/tsung/blob/68e509fedc703a7a2a444c90a8f2df882d517d8a/src/tsung/ts_mon_cache.erl#L165 ints_mon_cache:handle
.Any idea how we could prevent this to make the result data consistent, @nniclausse?