[
https://issues.apache.org/jira/browse/FLINK-5718?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15854677#comment-15854677
]
ASF GitHub Bot commented on FLINK-5718:
---------------------------------------
Github user greghogan commented on a diff in the pull request:
https://github.com/apache/flink/pull/3276#discussion_r99674991
--- Diff: docs/setup/config.md ---
@@ -86,7 +86,7 @@ The default fraction for managed memory can be adjusted
using the `taskmanager.m
- `taskmanager.memory.segment-size`: The size of memory buffers used by
the memory manager and the network stack in bytes (DEFAULT: 32768 (= 32
KiBytes)).
-- `taskmanager.memory.preallocate`: Can be either of `true` or `false`.
Specifies whether task managers should allocate all managed memory when
starting up. (DEFAULT: false). When `taskmanager.memory.off-heap` is set to
`true`, then it is advised that this configuration is also set to `true`. If
this configuration is set to `false` cleaning up of the allocated offheap
memory happens only when the configured JVM parameter MaxDirectMemorySize is
reached by triggering a full GC.
+- `taskmanager.memory.preallocate`: Can be either of `true` or `false`.
Specifies whether task managers should allocate all managed memory when
starting up. (DEFAULT: false). When `taskmanager.memory.off-heap` is set to
`true`, then it is advised that this configuration is also set to `true`. If
this configuration is set to `false` cleaning up of the allocated offheap
memory happens only when the configured JVM parameter MaxDirectMemorySize is
reached by triggering a full GC. **Note:** For streaming setups, we highly
recommend to set this value to `false` as the core state backends currently do
not use the managed memory.
--- End diff --
Should this warning also be added to `flink-conf.yaml`?
> Handle JVM Fatal Exceptions in Tasks
> ------------------------------------
>
> Key: FLINK-5718
> URL: https://issues.apache.org/jira/browse/FLINK-5718
> Project: Flink
> Issue Type: Improvement
> Components: Local Runtime
> Reporter: Stephan Ewen
> Assignee: Stephan Ewen
>
> The TaskManager catches and handles all types of exceptions right now (all
> {{Throwables}}). The intention behind that is:
> - Many {{Error}} subclasses are recoverable for the TaskManagers, such as
> failure to load/link user code
> - We want to give eager notifications to the JobManager in case something
> in a task goes wrong.
> However, there are some exceptions which should probably simply terminate the
> JVM, if caught in the task thread, because they may leave the JVM in a
> dysfunctional limbo state:
> - {{OutOfMemoryError}}
> - {{InternalError}}
> - {{UnknownError}}
> - {{ZipError}}
> These are basically the subclasses of {{VirtualMachineError}}, except for
> {{StackOverflowError}}, which is recoverable and usually recovered already by
> the time the exception has been thrown and the stack unwound.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
