[
https://issues.apache.org/jira/browse/HADOOP-7983?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eli Collins updated HADOOP-7983:
--------------------------------
Attachment: hadoop-7983.txt
Patch attached. It depends on HADOOP-7938.
The fencers now get passed the target service as an argument at runtime
(instead of not specific or configured statically). For the ssh-based fencer
this means that the hostname we ssh to is the one passed to the failover
command (HDFS-2808 will allow this to be a logical name but it will get mapped
to the service address). It's possible that admins could want to use a
separate interface/hostname for ssh'ing than the service address. To handle
this we could allow admins to configure node-specific fencing methods or a
config that map service addresses to ssh hostnames. I think we can punt on this
for now unless it's needed in practice.
> HA: failover should be able to pass args to fencers
> ---------------------------------------------------
>
> Key: HADOOP-7983
> URL: https://issues.apache.org/jira/browse/HADOOP-7983
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: ha
> Affects Versions: HA Branch (HDFS-1623)
> Reporter: Eli Collins
> Assignee: Eli Collins
> Attachments: hadoop-7983.txt
>
>
> Currently fencing method args are passed in the config (eg
> "sshfence(host1,8022)" indicates to fence the service running on port 8022 on
> host1. The target service to fence should be determined by the failover (we
> fence the currently active service) not configured statically.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
