accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Josh Elser (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-925) Launch scripts should use a PIDfile
Date Mon, 29 Jun 2015 21:35:04 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-925?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14606449#comment-14606449
] 

Josh Elser commented on ACCUMULO-925:
-------------------------------------

Somewhere [~mdrob] brought up the point of looking at what's been done upstream in Hadoop-3
via HADOOP-9902

I noticed that I can easily add in LSB-compatible/recommended return codes -- will include
this in a v3. Billie already made a nice {{ACCUMULO_IDENT_STR}} that can be provided by users
which is good and called out.

The other big change I see is the removal of {{\*\-daemon.sh}} scripts in favor of using {{--daemon}}.
For example, previously {{hadoop-daemon.sh start datanode}} would start a backgrounded datanode
on the local node, using Hadoop 3 the command would be {{hdfs --daemon start datanode}}. I
don't think we want to try to address that in this patch -- we're gaining the PID support
and some good separation around setup code and code to actually launch the process (helps
us separate what must be run on a remote node). Moving towards a {{*-daemon.sh}} script was
something I was thinking would be good, but I think matching upstream hadoop would be a better
goal. Either way, not intending to do it here.

> Launch scripts should use a PIDfile
> -----------------------------------
>
>                 Key: ACCUMULO-925
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-925
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: scripts
>            Reporter: Christopher Tubbs
>            Assignee: Billie Rinaldi
>             Fix For: 1.8.0
>
>         Attachments: ACCUMULO-925.1.patch, ACCUMULO-925.2.patch
>
>
> Start scripts should create PIDfiles to store the PID of running processes in a well
known location (example: /var/run/accumulo/tserver.pid or $ACCUMULO_HOME/tserver.pid), for
the following benefits:
> # Identify running services on a machine without executing and parsing the system process
list, so stop scripts can kill them when they are unresponsive.
> # Prevent multiple instances of the same application from starting up (an environment
variable for the location of the PIDfile can be used to allow multiple instances if it is
desirable to do so).
> # Potentially provide an alternate mechanism for terminating a process by deleting its
PIDfile rather than its lock in Zookeeper.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message