ElasticSearch Native Client
Services like Bernard, Ernest and Wally query the ElasticSearchServer via
ElasticSearchClient to get Lisitng Level Data, Offer-Mappings & Victor-
References.
Bernard Client
Ernest Client
Wally Client
ElasticSearch
Client
ElasticSearch
Server
Bernard - Offer Computation against a listing
Ernest - Offer Creation on Seller Dashboard
Wally - Scheduler for offer computation & Redriver
santa-prodparallell-0003.nm.flipkart.com:9200/_plugin/head/
There is only one Cluster of ElasticSearch and all the
three DSP-Santa services query to it. In case it goes
down or cluster health is affected, we need to identify
which service is causing the problem so that we can
furthur investigate into the particular service's server.
Solution
What we could do is we can write a wrapper to the ElasticSearch Client and all the services will
use this wrapper instead of using the ES-Client directly. The wrapper will invoke the ES-client with
all the calculations around the function calls. With the call-time data we could generate metrics and
publish it on the JMX port of the client. From there the data will be pushed to services like Cosmos
which will draw graphs around the data.
For synchronous functions, time is taken across the function calls. For asynchronous functions, the
original ActionListener is decorated with yammer metrics and passed on to the original ES-Client.
ElasticSearch Wrapper Client
Bernard Client
Ernest Client
Wally Client
ElasticSearch
Client
ElasticSearch
Server
ElasticSearch
Wrapper
Yammer
Metrics
JMX
port
collect
daemon
COSMOS
What we actually did?
Taking help of TImer from Yammer Library, we measured the frequency of, and time taken, by the
queries for each client. The metrics are published on the JMX port of the client, from where the
Collect-D collects it and publishes it on COSMOS. Cosmos will represent all data graphically and will
help to analyse and debug is anything goes wrong. Basically 2 use-cases had to be covered.
Synchronous
Functions
Asynchronous
Functions
We measure time spent across function calls.
That is same as time for each query.
To measure time for each query , we check when the
ActionListner detects an Event or gives an Exception
Project : ElasticSearch Wrapper/
ElasticSearch Metrics
Generator
Project Manager : Mrs. Vinita Bansal
Project Mentor : Mr. Prarabdh Awasthi
- Dhruv Chandhok

More Related Content

PPT
Amazon’s Cloud Computing Efforts
PDF
Introducing Athena: 08/19 Big Data Application Meetup, Talk #3
PDF
Kubernetes at Telekom Austria Group
PPTX
03 spring cloud eureka service discovery
PPT
awstalk.ppt
PDF
Project Frankenstein: A multitenant, horizontally scalable Prometheus as a se...
PDF
MongoDB .local Bengaluru 2019: Distributed Transactions: With Great Power Com...
PPTX
Scalable Google Cloud Payroll Project- Presentation
Amazon’s Cloud Computing Efforts
Introducing Athena: 08/19 Big Data Application Meetup, Talk #3
Kubernetes at Telekom Austria Group
03 spring cloud eureka service discovery
awstalk.ppt
Project Frankenstein: A multitenant, horizontally scalable Prometheus as a se...
MongoDB .local Bengaluru 2019: Distributed Transactions: With Great Power Com...
Scalable Google Cloud Payroll Project- Presentation

What's hot (9)

PDF
MQTT - REST Bridge using the Smart Object API
PDF
Ceilometer + Heat = Alarming
PDF
OpenStack Ceilometer
PDF
Softwerkskammer Lübeck 08/2018 Event Sourcing and CQRS
PPTX
Rack middlewares - ins outs
PDF
From Ceilometer to Telemetry: not so alarming!
PPTX
Web Real-time Communications
PDF
Clustering and load balancing : jboss
PPTX
Adopting serverless
MQTT - REST Bridge using the Smart Object API
Ceilometer + Heat = Alarming
OpenStack Ceilometer
Softwerkskammer Lübeck 08/2018 Event Sourcing and CQRS
Rack middlewares - ins outs
From Ceilometer to Telemetry: not so alarming!
Web Real-time Communications
Clustering and load balancing : jboss
Adopting serverless
Ad

Similar to ELASTICSEARCH-CLIENT (20)

PDF
Monitoring as Software Validation
PDF
SHOWDOWN: Threat Stack vs. Red Hat AuditD
PDF
Application metrics - Confoo 2019
PDF
Q radar architecture deep dive
PPTX
Application Timeline Server Past, Present and Future
PDF
Barga IC2E & IoTDI'16 Keynote
TXT
Application enumeration and launch
PDF
Application Metrics (with Prometheus examples)
PDF
The present and future of Serverless observability
PDF
The present and future of serverless observability (QCon London)
PDF
The present and future of Serverless observability
PPTX
Vulnerability Discovery in the Cloud
PDF
QConSF 2014 talk on Netflix Mantis, a stream processing system
PPTX
Cinema booking system | Movie Booking System
PPTX
Accelerating analytics on the Sensor and IoT Data.
PDF
Introduction to Marionette Collective
PPTX
Samantha Wang [InfluxData] | Best Practices on How to Transform Your Data Usi...
PDF
Choose Right Stream Storage: Amazon Kinesis Data Streams vs MSK
PDF
Application Metrics (with Prometheus examples) #PHPDD18
PPTX
Fabric - Realtime stream processing framework
Monitoring as Software Validation
SHOWDOWN: Threat Stack vs. Red Hat AuditD
Application metrics - Confoo 2019
Q radar architecture deep dive
Application Timeline Server Past, Present and Future
Barga IC2E & IoTDI'16 Keynote
Application enumeration and launch
Application Metrics (with Prometheus examples)
The present and future of Serverless observability
The present and future of serverless observability (QCon London)
The present and future of Serverless observability
Vulnerability Discovery in the Cloud
QConSF 2014 talk on Netflix Mantis, a stream processing system
Cinema booking system | Movie Booking System
Accelerating analytics on the Sensor and IoT Data.
Introduction to Marionette Collective
Samantha Wang [InfluxData] | Best Practices on How to Transform Your Data Usi...
Choose Right Stream Storage: Amazon Kinesis Data Streams vs MSK
Application Metrics (with Prometheus examples) #PHPDD18
Fabric - Realtime stream processing framework
Ad

ELASTICSEARCH-CLIENT

  • 1. ElasticSearch Native Client Services like Bernard, Ernest and Wally query the ElasticSearchServer via ElasticSearchClient to get Lisitng Level Data, Offer-Mappings & Victor- References. Bernard Client Ernest Client Wally Client ElasticSearch Client ElasticSearch Server Bernard - Offer Computation against a listing Ernest - Offer Creation on Seller Dashboard Wally - Scheduler for offer computation & Redriver santa-prodparallell-0003.nm.flipkart.com:9200/_plugin/head/ There is only one Cluster of ElasticSearch and all the three DSP-Santa services query to it. In case it goes down or cluster health is affected, we need to identify which service is causing the problem so that we can furthur investigate into the particular service's server. Solution What we could do is we can write a wrapper to the ElasticSearch Client and all the services will use this wrapper instead of using the ES-Client directly. The wrapper will invoke the ES-client with all the calculations around the function calls. With the call-time data we could generate metrics and publish it on the JMX port of the client. From there the data will be pushed to services like Cosmos which will draw graphs around the data. For synchronous functions, time is taken across the function calls. For asynchronous functions, the original ActionListener is decorated with yammer metrics and passed on to the original ES-Client.
  • 2. ElasticSearch Wrapper Client Bernard Client Ernest Client Wally Client ElasticSearch Client ElasticSearch Server ElasticSearch Wrapper Yammer Metrics JMX port collect daemon COSMOS What we actually did? Taking help of TImer from Yammer Library, we measured the frequency of, and time taken, by the queries for each client. The metrics are published on the JMX port of the client, from where the Collect-D collects it and publishes it on COSMOS. Cosmos will represent all data graphically and will help to analyse and debug is anything goes wrong. Basically 2 use-cases had to be covered. Synchronous Functions Asynchronous Functions We measure time spent across function calls. That is same as time for each query. To measure time for each query , we check when the ActionListner detects an Event or gives an Exception Project : ElasticSearch Wrapper/ ElasticSearch Metrics Generator Project Manager : Mrs. Vinita Bansal Project Mentor : Mr. Prarabdh Awasthi - Dhruv Chandhok