nastel / jesl Goto Github PK

Upgrade to httpclient 4.5.x to address security vulnerabilities.
https://github.com/Nastel/JESL/network/alert/pom.xml/org.apache.httpcomponents:httpclient/open

Using the custom version of order-process.xml simulation, running 5000 iterations, it appears that not all generated tracking items are streamed to jKool server. The metrics printed at the end of the simulation run appear to show this.

Since each iteration of my custom order-process simulation generates 7 activities and 18 events, the total number of items that should be reported to jKool server should be:

 7 * 5000 = 35,000 activities
18 * 5000 = 90,000 events

The following metrics reported by simulator confirm this:

TrackerImpl/tracker-activities:                             35,000
TrackerImpl/tracker-events:                                 90,000

However, the number of items processed by the sink are:

SimulatedEventSink/sink-activities:                         35,000
SimulatedEventSink/sink-events:                             88,020
SimulatedEventSink/sink-skipped:                             1,980
JKCloudEventSink/sink-direct-writes:                       123,020
JKCloudEventSink/sink-sent-messages:                       123,020

No reason is given as to why the events are being skipped

Currently, the streaming access token is only sent once, when connection is established.

Request is to include the token in EVERY HTTP message sent. Suggestion is to use the Authorization header.

This is for future use, where routing and throttling would be done by an external gateway.

Running the order-process.xml simulation, events sporatically get "lost". The simulator sink results show the correct number of events sent, but the jKool Gateway does not received that many. When it does happen, it is always one of the SQL.Execute events in the AcceptOrder activity (first activity in simulation). I've seen it running both a single iteration and several iterations.

This was testing the latest code from both jesl and tnt4j repositories.

We have a job that runs every 5 minutes and will generate a few hundred records to be posted each time around. Is it possible to post them all with one POST request? We can generate a separate post request for each, it's just a lot more overhead. Also, batching many records seems a pretty common use case.