Rui Fan created FLINK-34104:
-------------------------------
Summary: Improve the ScalingReport format of autoscaling
Key: FLINK-34104
URL: https://issues.apache.org/jira/browse/FLINK-34104
Project: Flink
Issue Type: Improvement
Components: Autoscaler
Reporter: Rui Fan
Assignee: Rui Fan
Fix For: kubernetes-operator-1.8.0
Currently, the scaling report format is
{color:#6a8759} Vertex ID %s | Parallelism %d -> %d | Processing capacity %.2f
-> %.2f | Target data rate %.2f{color}
{color:#172b4d}It has 2 disadvantages:{color}
# {color:#172b4d}When one job has multiple vertices, the report of all
vertices are mixed together without any separator{color}{color:#172b4d}, here
is an example:{color}
** {color:#172b4d}Scaling execution enabled, begin scaling vertices: Vertex ID
ea632d67b7d595e5b851708ae9ad79d6 | Parallelism 2 -> 1 | Processing capacity
800466.67 -> 320186.67 | Target data rate 715.10 Vertex ID
bc764cd8ddf7a0cff126f51c16239658 | Parallelism 2 -> 1 | Processing capacity
79252.08 -> 31700.83 | Target data rate 895.93 Vertex ID
0a448493b4782967b150582570326227 | Parallelism 8 -> 16 | Processing capacity
716.05 -> 1141.00 | Target data rate 715.54{color}
** {color:#172b4d}We can see the Vertex ID is the beginning of each vertex
report, it doesn't have any {color}{color:#172b4d}separator with the last
vertex.
{color}
# {color:#172b4d}This format is non-standard{color}{color:#172b4d}, it's hard
to deserialize.{color}
** {color:#172b4d}When job enables the autoscaler and disable the
scaling.{color}
** {color:#172b4d}Flink platform maintainer wants to show the scaling report
in WebUI, it's helpful to using the report result for flink users.{color}
** {color:#172b4d}So easy to deserialize is useful for these flink
platform.{color}
h2. {color:#172b4d}Solution:{color}
{color:#172b4d}Serializing the scaling report with json format, it's easy to
read and deserialize.{color}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
