Logs & Application Improvement

• Merchant & Token in Logs
• Call Api unwrapping error (Fix for sending 400 to frontend upon bad request and 401 only upon auth failed)
• Added request timeout middleware (To check if it can fix 0 DC)
• Added Drainer lag metrics in Prometheus
• Added speed in Kafka sent by frontend
• Removed Panics to fix 0 DC errors
• Reset drainer memory after every drain, to check if it can fix the sudden increase in Pods memory and then memory not being release issue.

Reduced Repeated Drainer Code & Fixed Prometheus External API Call Metrics

Reduced Repeated Drainer Code & Fixed Prometheus External API Call Me…

…trics

On Ride Location Updates Flow Fix & Prom Metrics Addition

v1.0.4

On Ride Location Updates Flow Fix & Prom Metrics Addition

Added Prometheus Metrics for Drainer Lag

Added Prometheus Metrics for calculating Drainer Lag for batching and draining location updates to Redis in one call.

Location Updates Filtering Fixes and Minor Improvements

• Added warn and error logs for every failure case, even when we ignore.
• Filtering logic for accuracy moved to only OnRide location updates.
• Fixed correct Driver Mode passed to Kafka.
• Current bucket based on min(location_ts, curr_ts). If future timestamp, then we change it to current with a Warning log.
• Set driver last known location also based on min(location_ts, curr_ts).
• Removed redundant code repetition from updateLocation handler.

Bug Fixes and Enhancements

• We were throwing 5xx in driver location tracking when called after rideEnd - To solve this we added proper error codes and mapped it to 400 so that we can get distinction based on error code in grafana.
• Location Unserviceable issue as earlier we were only taking location from 0th index - To solve this, Now we are sorting location based on timestamp and filtering based on Accuracy first and then only taking the latest location lat long for validating City.
• In case where content is large or request body is not of Json type. The error is mapped to (422 - UNPROCESSABLE_ENTITY), didn't map it to 416 as it means CONTENT_TOO_LARGE which is not Generic response.
• Pipelined msetnx for setting lock on redis to fix the issue where lock is set and pod is terminated Ungracefully so expiry is not set on those keys and the driver is blocked for location update with Hitlimiexceeded error.
• Prom Metrics correction for AB to work.
• Bug Fix, for Driver Location Accumulation resulting in increase Snap to Road calls.