Skip to main content

Suggest an Idea

Vote for your favorite Qlik product ideas and add your own suggestions.

Announcements
Welcome to Qlik Community! Check out our new navigation! FIND OUT MORE

Qlik Replicate windows server was rebooted abruptly

SaiKoya
Contributor II
Contributor II

Qlik Replicate windows server was rebooted abruptly

Problem Statement: Due to abrupt restart of replicate services or windows server reboot (planned vs unplanned), the replicate task writing to databricks endpoint is abruptly ended and as a result, partial or corrupt seq files are being written to databricks tables. When windows server comes up and Replicate services are started, replicate task continues to run from where it left. The problem/behavior we currently see is, Replicate task continues to process and append new seq files to the databricks table WITHOUT COMPLAINING.

 

Impact: As replicate is NOT complaining for any issues, the task continues to load/process new data, until our Business Teams or Compose Jobs start failing. At that point, we need to identify “manually” via business inputs or compose team inputs each and every table that was corrupted and rectify the data by cleaning the corrupt files/partitions and then process the data from the point of corruption.

 

Ask: While we understand that databricks is a file based target and abrupt shutdown of replicate services or windows server can result in corruption, we need Replicate to TELL US or identify the corrupted tasks so that data processing can be STOPPED from the time of corruption. At this point we are completely relying on our business or compose teams to let us know the data corruption and hence want product to handle this instead of relying on downstreams.

 

Possible Solutions: Due to abrupt shutdown, replicate might or might not have received the “completion signal” from databricks for the file it was processing. Replicate stores this information internally and upon task initialization, do NOT allow the task to be resumed and spits out a message saying ‘task stopped abnormally and might have corrupt tables’. This will at least narrow down to a few tables/tasks. But again, the ask is to get full information via the product!!