You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We need this feature to scale. Here's valid case: there's limit for each partition on each consumer group. If there are more than 5 concurrent jobs in a spark app, we are getting the exception - "Exceeded the maximum number of allowed receivers per partition in a consumer group which is 5."
Since we can't add more concurrent jobs, in order to perform faster, the solution is to create more consumer groups, and create multiple spark apps, each would hit different consumer group on specific partition(s). For example, a 32 partitions of Eventhub (1 default consumer group) read by 5 concurrent spark jobs would be much slower than 16 consumer groups, each assigned 2 partitions and read by 1 Spark app. So we can have 16 Spark apps, each has 5 concurrent jobs - total would be 80 concurrent jobs. 16x faster !!
The text was updated successfully, but these errors were encountered:
follow up on #220
We need this feature to scale. Here's valid case: there's limit for each partition on each consumer group. If there are more than 5 concurrent jobs in a spark app, we are getting the exception - "Exceeded the maximum number of allowed receivers per partition in a consumer group which is 5."
Since we can't add more concurrent jobs, in order to perform faster, the solution is to create more consumer groups, and create multiple spark apps, each would hit different consumer group on specific partition(s). For example, a 32 partitions of Eventhub (1 default consumer group) read by 5 concurrent spark jobs would be much slower than 16 consumer groups, each assigned 2 partitions and read by 1 Spark app. So we can have 16 Spark apps, each has 5 concurrent jobs - total would be 80 concurrent jobs. 16x faster !!
The text was updated successfully, but these errors were encountered: