Hi Thiago, Thanks for checking out Metron! Great to hear you're trying out MaaS, we love the feedback. I'm not sure what the expected response time should be for your model - that's probably worth a separate look (Casey Stella - any thoughts here?). If your model is taking 5-10 seconds to return a result, it's also possible you may be hitting timeouts in Storm and Kafka for messages in your enrichment topology. Here are a couple properties you might be interested in:
1. https://github.com/apache/storm/blob/master/conf/defaults.yaml#L258 - I think you'll have to add the property to remote.yaml file for enrichment find that in $METRON_HOME. 2. session.timeout.ms - https://github.com/apache/metron/blob/master/metron-platform/Performance-tuning-guide.md#enrichment-tuning The system is at-least-once processing. If the message is timing out with Storm, you should also see an error for the tuple in the Storm UI. Best, Mike Miklavcic On Thu, Sep 12, 2019 at 12:06 PM Thiago Rahal Disposti < [email protected]> wrote: > Good evening, everyone! > > I am currently trying to make MaaS work for an AI model that I have > created, capable of predicting wheter a webpage that passed through squid > is phishing. I managed to make MaaS work for the DGA model ( > https://metron.apache.org/current-book/metron-analytics/metron-maas-service/index.html), > however when it comes to my own model I sometimes get repeated events > indexed in ElasticSearch (specially when I have 5+ requests consecutively). > I think that the problem is related with the fact that my model currently > takes 5-10 seconds to return an "is_malicious" result, while the DGA model > returns it almost immediately. > Has anyone experienced something similar? > > Thanks in advance, > Thiago Rahal >
