1996fanrui commented on code in PR #23247:
URL: https://github.com/apache/flink/pull/23247#discussion_r1423360642
##########
flink-core/src/main/java/org/apache/flink/configuration/RestartStrategyOptions.java:
##########
@@ -184,7 +184,7 @@ public class RestartStrategyOptions {
public static final ConfigOption<Double>
RESTART_STRATEGY_EXPONENTIAL_DELAY_BACKOFF_MULTIPLIER =
ConfigOptions.key("restart-strategy.exponential-delay.backoff-multiplier")
.doubleType()
- .defaultValue(2.0)
+ .defaultValue(1.2)
Review Comment:
Thanks @mxm for the feedback!
`1.5` is fine for me, and I'd like to cc @zhuzhurk who propose change these
2 default value as well!
Max and Mason have a little feedback after voting in the [user mail
list](https://lists.apache.org/thread/6glz0d57r8gtpzq4f71vf9066c5x6nyw), and
Max and I had a offline discussion yesterday. Max think the 1.2 is a little
small or aggressive. Here is the reason:
- Every time the job restarts, it will make a bunch of calls to the
Kubernetes API, e.g. read/write to config maps, create task managers.
- When his producation had the default fixed-delay(1s) restart strategy
turned on. A Kubernetes cluster became instable.
Following is the relationship between restart-attempts and retry-delay-time:
- The `delay-time` will reach 1 min after 12 attempts when
`backoff-multiplier` is 1.5
- The `delay-time` will reach 1 min after 24 attempts when
`backoff-multiplier` is 1.2
Hey @zhuzhurk , what do you think about setting`1.5` as the default value?
If you agree it, I will update the FLIP and feedback it to the [user mail
list](https://lists.apache.org/thread/6glz0d57r8gtpzq4f71vf9066c5x6nyw), and go
ahead this PR.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@flink.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
