We had a legacy .NET web-service that uses WSE 3.0. We are the service provider and we have an intermediate authentication server as well as the consumer. We had intermittent mysterious socket time out exceptions.
A thorough network snoop was conducted across 4 network teams (we span across 2 intranet and 2 WAN domains). Everything was healthy with <10ms latency between hops.
It was puzzling and a stroke of genius came by and we decided to check the inputtrace.webinfo and outputtrace.webinfo (default logging configs). Nothing special about it except for the size. Monthly scheduled jobs archival results in file size in excess of 600mb. Some team members had argued that logging mechanism should be asynchronous and should not have any influence on the response SOAP. There's no harm trying to flush out the logs and archive and do a re-test. Sure enough everything was back to < 0.5s
Our corrective action was to increase the frequency of the archival from monthly to weekly.
No comments:
Post a Comment