You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Matthias J. Sax (JIRA)" <ji...@apache.org> on 2015/08/28 03:17:45 UTC
[jira] [Updated] (FLINK-2588) Implement KillOptions
[ https://issues.apache.org/jira/browse/FLINK-2588?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Matthias J. Sax updated FLINK-2588:
-----------------------------------
Description:
In Storm, topologies run forever and need to be terminate via `LocalCluster.killTopology()` or `Client.killTopology()`. For both calls, it is possible to specify a `KillOptions` parameter. Currently, this parameter is ignored.
`KillOptions` in Storm offer only a single option:; `.set_wait_secs(int)`. This delays the killing by the specified number of seconds. (Need to double check if the call blocks for the specified number of seconds or not).
Furthermore, in Flink `killTopology()` is implemented to send the "stop" signal. As the "stop" signal is no hard termination, it would be nice to be able to send the "cancel" signal, too.
# For this, `FlinkKillOptions` should be introduced (extending `KillOptions`). With `FlinkKillOptions` it should be possible to set the signal type to "cancel".
# Additionally, an option should be available to make this call blocking (for "stop" and "cancel"); ie, the call returns not before the job is cleaned up.
# Last but not least, a "stop-wait-cancel" feature can be implements: send a "stop" signal and wait for a specific time. If job does not finish within this time, we assume that the stop signal was ignored and thus, a "cancel" is sent.
Require https://issues.apache.org/jira/browse/FLINK-2111 and https://issues.apache.org/jira/browse/FLINK-2338 to be resolved.
was:
In Storm, topologies run forever and need to be terminate via `LocalCluster.killTopology()` or `Client.killTopology()`. For both calls, it is possible to specify a `KillOptions` parameter. Currently, this parameter is ignored.
`KillOptions` in Storm offer only a single option:; `.set_wait_secs(int)`. This delays the killing by the specified number of seconds. (Need to double check if the call blocks for the specified number of seconds or not).
Furthermore, in Flink `killTopology()` is implemented to send the "stop" signal. As the "stop" signal is no hard termination, it would be nice to be able to send the "cancel" signal, too.
# For this, `FlinkKillOptions` should be introduced (extending `KillOptions`). With `FlinkKillOptions` it should be possible to set the signal type to "cancel".
# Additionally, an option should be available to make this call blocking (for "stop" and "cancel"); ie, the call returns not before the job is cleaned up.
# Last but not least, a "stop-wait-cancel" feature can be implements: send a "stop" signal and wait for a specific time. If job does not finish within this time, we assume that the stop signal was ignored and thus, a "cancel" is sent.
> Implement KillOptions
> ---------------------
>
> Key: FLINK-2588
> URL: https://issues.apache.org/jira/browse/FLINK-2588
> Project: Flink
> Issue Type: Improvement
> Components: Storm Compatibility
> Reporter: Matthias J. Sax
> Priority: Minor
>
> In Storm, topologies run forever and need to be terminate via `LocalCluster.killTopology()` or `Client.killTopology()`. For both calls, it is possible to specify a `KillOptions` parameter. Currently, this parameter is ignored.
> `KillOptions` in Storm offer only a single option:; `.set_wait_secs(int)`. This delays the killing by the specified number of seconds. (Need to double check if the call blocks for the specified number of seconds or not).
> Furthermore, in Flink `killTopology()` is implemented to send the "stop" signal. As the "stop" signal is no hard termination, it would be nice to be able to send the "cancel" signal, too.
> # For this, `FlinkKillOptions` should be introduced (extending `KillOptions`). With `FlinkKillOptions` it should be possible to set the signal type to "cancel".
> # Additionally, an option should be available to make this call blocking (for "stop" and "cancel"); ie, the call returns not before the job is cleaned up.
> # Last but not least, a "stop-wait-cancel" feature can be implements: send a "stop" signal and wait for a specific time. If job does not finish within this time, we assume that the stop signal was ignored and thus, a "cancel" is sent.
> Require https://issues.apache.org/jira/browse/FLINK-2111 and https://issues.apache.org/jira/browse/FLINK-2338 to be resolved.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)