[
https://issues.apache.org/jira/browse/FLINK-3172?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15058901#comment-15058901
]
ASF GitHub Bot commented on FLINK-3172:
---------------------------------------
GitHub user uce opened a pull request:
https://github.com/apache/flink/pull/1458
[FLINK-3172] Specify JobManager port in HA mode
- Adds support for job manager port configuration for standalone HA clusters
- Re-uses everthing introduced in #1416, but moves the `retry` method to
the JobManager and adds an optional sleep time between retries.
- Configuration via `recovery.jobmanager.port` (takes single ports, ranges,
or mix)
- This doesn't work with standalone clusters. It would be possible to
activate, but then the task managers will have to try each port in the range to
connect to a job manager (because in standalone mode there is no ZooKeeper
invovled)
---
Starting multiple job managers locally with the default configuration
(none) works as expected by picking random ports:
```bash
$ grep ".*Remoting started; listening on addresses.*" -r log
jobmanager-0.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:51132]
jobmanager-1.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:51131]
jobmanager-2.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:51134]
jobmanager-3.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:51136]
jobmanager-4.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:51133]
jobmanager-5.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:51135]
```
After configuring a port range via `recovery.jobmanager.port: 6123-6133`:
```bash
jobmanager-0.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:6125]
jobmanager-1.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:6123]
jobmanager-2.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:6124]
jobmanager-3.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:6128]
jobmanager-4.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:6127]
jobmanager-5.log:...Remoting started; listening on addresses
:[akka.tcp://[email protected]:6126]
```
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/uce/flink 3172-ha_jm_port
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/1458.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #1458
----
commit 61f33404dfd2ce2f8af6ddf8f90ab4b689c0d982
Author: Ufuk Celebi <[email protected]>
Date: 2015-12-15T19:39:17Z
[FLINK-3074] [yarn] Fix port range retry termination condition
commit 3c0708b155aefaf89e666c9cdcc0aa34473abc13
Author: Ufuk Celebi <[email protected]>
Date: 2015-12-15T19:42:06Z
[FLINK-3073] [dist] Fix JobManager command line argument
Removed streaming mode lead to wrong arguments being passed.
commit 7e2417cd1de040bf2d0e6736353fe00d38ce5cbc
Author: Ufuk Celebi <[email protected]>
Date: 2015-12-15T21:01:17Z
[FLINK-3172] [core, runtime, yarn] Allow port range for job manager with
high availability
----
> Specify jobmanager port in HA mode
> ----------------------------------
>
> Key: FLINK-3172
> URL: https://issues.apache.org/jira/browse/FLINK-3172
> Project: Flink
> Issue Type: Improvement
> Components: Distributed Runtime
> Affects Versions: 0.10.1
> Reporter: Ufuk Celebi
> Assignee: Ufuk Celebi
> Priority: Minor
>
> In HA mode, the job manager port is picked up randomly. In firewalled setups
> this can be problematic. We should add a way to use HA mode without random
> ports (like the web frontend currently does).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
