saltstack-formulas / accumulo-formula Goto Github PK

It runs zkCli.sh and dumps its output to a temporary file, then greps that file for "accumulo" (specifically looking for the /accumulo directory in ZooKeeper). But, if the ZooKeeper host has "accumulo" in its name, this gets included in a log message, and so the state concludes that Accumulo has already been set up even if it hasn't.

Here's the output from a successful run:

dmaze@devhub2:~$ salt dmaze-dg-perf-acc00 cmd.run 'cat /tmp/acc.status'
dmaze-dg-perf-acc00:
    2014-09-06 01:05:23,177 [myid:] - INFO  [main:ZooKeeper@438] - Initiating client connection, connectString=dmaze-dg-perf-acc00:2181 sessionTimeout=30000 watcher=org.apache.zookeeper.ZooKeeperMain$MyWatcher@637a7476

If instead you had named the ZK host something like "accumulo-master", this would show up in that connectString parameter, and the grep would incorrectly succeed.

https://accumulo.apache.org has no install time dependency on hadoop/zookeeper.
This formula should be fixed to focus on accumulo and get it installed.

Hard dependencies on external formula should be isolated and off by default.

• hadoop/zk are runtime dependencies only.
• uses should be allowed to install hadoop/zk in any manner (do not enforce salt).

jinja2.exceptions.TemplateNotFound: hadoop/settings.sls
; line 6
---
{%- if grains['os_family'] in ['RedHat'] %}
redhat-lsb-core:
  pkg.installed
{% endif %}

{%- from 'hadoop/settings.sls' import hadoop with context %}    <======================
{%- from 'hadoop/hdfs/settings.sls' import hdfs with context %}
{%- from 'zookeeper/settings.sls' import zk with context %}
{%- from 'accumulo/settings.sls' import accumulo with context %}

Add a "manual profile" that allows the setup of the values that so far were encapsulated in the memory profiles. This is intending for adjusting/tuning accumulo instances beyond basic needs.