cookbook 'elkstack', '= 2.1.4'
elkstack
(29) Versions
2.1.4
-
Follow36
Installs/Configures elkstack
cookbook 'elkstack', '= 2.1.4', :supermarket
knife supermarket install elkstack
knife supermarket download elkstack
elkstack
Elasticsearch, Logstash, and Kibana stack. Due to the recommendations of the
community, we are not using the embedded elasticsearch functionality of logstash
at this point. This cookbook provides recipes for all three components, along
with wrapper recipes such as single
or cluster
to facilitate different use
cases.
This stack's design is intended for one or many standalone nodes, with a full
stack of elasticsearch, logstash, and kibana. The only difference between one
and many nodes is that elasticsearch is clustered together. Data dispatched to
Logstash on a particular node will use the local elasticsearch transport
interface to index those logs to the node (and thus, the cluster). HTTP traffic
dispatched to Kibana on port 80 on any node will also use the local
elasticsearch HTTP interface to fetch and manipulate data.
Please read the individual recipe summaries to understand what each recipe does,
as well as what each wrapper recipe is actually wrapping. As much as possible,
upstream attributes have been exposed/overriden for our needs.
Things you should know
Please note that this cookbook does not restart elasticsearch automatically,
in order to avoid causing an outage of the cluster. It does restart nginx and
logstash, however. You will have to restart elasticsearch after the initial
bootstrap. You may also need to bounce logstash if it seems confused about
losing a connection to eleasticsearch (unusual, but happens).You may want to consider adjusting
node['elasticsearch']['discovery']['search_query']
if you are sharing one cluster among multiple environments. Just put a chef
search in that attribute and this will use that search instead of one scoped to
chef environments.You may want to consider adjusting
node['elasticsearch']['allocated_memory']
if you are seeing an initial convergence failure (see #50).
The chef client has been known to take up to 500mb or more on initial
convergence. Combined with an initial allocation of 40% memory for ES, and 20%
for logstash, that only leaves about 40% for the OS and chef. On a 2gb server,
that ends up being 800mb for ES, about 400mb for logstash, leaving 800mb for
the OS and the initial chef client run. After the initial run, memory footprint
for the chef-client tends to be much, much lower, and ES is able to start.The agent recipes requires a pre-generated SSL key and certificate with
something likeopenssl req -x509 -newkey rsa:2048 -keyout lumberjack.key -out
. This key and certificate data should be
lumberjack.crt -nodes -days 1000
placed in data bag with namenode['elkstack']['config']['lumberjack_data_bag']
underkey
andcertificate
keys, and base64 encoded into a single line
string. You may also supply these with some other method and populate the
appropriatenode.run_state
values (see_secrets.rb
for more details). Note
that this is not a PKI trust model, but an
explicit trust model.
[Changelog](CHANGELOG.md)
See CHANGELOG.md for additional information about changes to this stack over time.
Supported Platforms
Ubuntu 12.04
Ubuntu 14.04
CentOS 6.5
Attributes
<table>
<tr>
<th>Key</th>
<th>Type</th>
<th>Description</th>
<th>Default</th>
</tr>
<tr>
<td><tt>['elkstack']['config']['logstash']['instance_name']</tt></td>
<td>String</td>
<td>Default logstash instance name</td>
<td><tt>server</tt></td>
</tr>
<tr>
<td><tt>['elkstack']['config']['cluster']</tt></td>
<td>Boolean</td>
<td>Whether to search for and connect Elasticsearch to cluster nodes</td>
<td><tt>false</tt></td>
</tr>
<tr>
<td><tt>['elasticsearch']['discovery']['search_query']</tt></td>
<td>String</td>
<td>A query to search for and connect Elasticsearch to cluster nodes</td>
<td>(see attributes/elasticsearch.rb
)</td>
</tr>
<tr>
<td><tt>['logstash_forwarder']['config']['files']</tt></td>
<td>Hash</td>
<td>See customizing the stack section below.</td>
<td>Most logs in /var/log
</td>
</tr>
<tr>
<td><tt>['elkstack']['config']['data_disk']['disk_config_type']</tt></td>
<td>Boolean or String</td>
<td>See customizing the stack section below.</td>
<td><tt>false</tt></td>
</tr>
<tr>
<td><tt>['elkstack']['iptables']['enabled']</tt></td>
<td>Boolean</td>
<td>Enable/Disable iptables functionality</td>
<td><tt>true</tt></td>
</tr>
<tr>
<td><tt>['elkstack']['config']['site_name']</tt></td>
<td>String</td>
<td>Control the name of the self-signed SSL key and cert in /etc/nginx/ssl</td>
<td><tt>kibana</tt></td>
</tr>
<tr>
<td><tt>['elkstack']['config']['kibana']['redirect']</tt></td>
<td>Boolean</td>
<td>Enable/Disable nginx redirect for kibana from port 80 to port 443</td>
<td><tt>true</tt></td>
</tr>
<tr>
<td><tt>node.run_state['elkstack_kibana_username']</tt> and <tt>['elkstack']['config']['kibana']['username']</tt></td>
<td>String</td>
<td>Default username for basic auth for kibana, run_state used first</td>
<td><tt>kibana</tt></td>
</tr>
<tr>
<td><tt>node.run_state['elkstack_kibana_password']</tt></td>
<td>String</td>
<td>Password for basic auth for kibana</td>
<td>random from <tt>Opscode::OpenSSL::Password</tt></td>
</tr>
<tr>
<td><tt>['elkstack']['config']['lumberjack_data_bag']</tt></td>
<td>String</td>
<td>Data bag name for lumberjack key and certificate</td>
<td><tt>lumberjack</tt></td>
</tr>
<tr>
<td><tt>['elkstack']['config']['additional_logstash_templates']</tt></td>
<td>Hash</td>
<td>Additional template files to be added by cookbooks depending on this one</td>
<td><tt>See attribute file for examples</tt></td>
</tr>
</table>
Customizing the stack
To override local storage for elasticsearch nodes (the stack will format and mount, as well as configure elasticsearch), set ['elkstack']['config']['data_disk']['disk_config_type']
to custom
and provide each storage device and mount point in the following way:
```ruby
disk_config = {
'file_system' => 'ext4',
'mount_options' => 'rw,user',
'mount_path' => '/usr/local/var/data/elasticsearch/disk1',
'format_command' => 'mkfs -t ext4 ',
'fs_check_command' => 'dumpe2fs'
}
node.override['elasticsearch']['data']['devices']['/dev/xvde1'] = disk_config
node.override['elasticsearch']['path']['data'] = disk_config['mount_path']
```
To add additional logstash configuration to this stack, simply add additional
templates in your wrapper cookbook. They should be placed in
"#{@basedir}/#{@instance}/etc/conf.d"
(see the config provider in the logstash
cookbook). If you choose to use logstash-forwarder instead of the regular agent,
please see the hash structure in attributes/forwarder.rb
for adding additional
files for the forwarder to watch and forward, node['logstash_forwarder']['config']['files']
.
To override the nginx configuration, simply supply a new template and specify
your cookbook using ['kibana']['nginx']['template_cookbook']
and
['kibana']['nginx']['template']
. You can also override just the password for
the reverse proxy using node.run_state['elkstack_password']
.
By default, platformstack will call the `::agent
recipe here. If you have a
need for the forwarder recipe instead, just please note that you should turn off
the platformstack flag for logging, and include the elkstack::agent
recipe
directly.
To override anything else, set the appropriate node hash (logstash
, kibana
, or elasticsearch
).
Usage
elkstack::default
Default recipe, does not do anything.
elkstack::single
A simple wrapper recipe that sets up Elasticsearch, Logstash, and Kibana. Also
configures an rsyslog sink into logstash on the local box. Everything except
Logstash and Kibana is locked down to listen only on localhost.
elkstack::cluster
A simple wrapper recipe that sets up Elasticsearch, Logstash, and Kibana. Also
configures an rsyslog sink into logstash on the local box. Sets the cluster flag
so that the elasticsearch recipe builds it in a cluster-friendly way.
elkstack::agent
A simple wrapper recipe that sets up a logstash agent on the local box. Also
configures an rsyslog sink into logstash on the local box.
elkstack::forwarder
A python-based alternative to the normal agent, configured simply to watch logs
forward them directly on to the cluster. This project is in heavy development,
and is not publishing releases very often, so the packaged versions may be quite
old or buggy. As of the addition of the recipe, the package was almost a year
behind current development, but only because there also had been no releases
either.
elkstack::elasticsearch
Leans on the upstream elasticsearch/cookbook-elasticsearch
cookbook for much
of its work. We do override the default set of plugins to be installed, as well
as the amount of JVM heap. See attributes/default.rb
for those settings.
This recipe also tags the node so that other nodes that run this recipe can
discover it, and configure Elasticsearch appropriately to join their cluster.
It uses a tag, the current chef environment, and the cluster name as the default
search criteria.
Most of this is configurable using the upstream Elasticsearch cookbook's
attributes, including the chef search itself. There is not an easy toggle to
turn off the search, however.
Enables iptables rules if default['elkstack']['iptables']['enabled'] not nil
elkstack::logstash
Leans on the upstream lusis/chef-logstash
cookbook for much
of its work. We do override the default set of plugins to be installed, as well
as the amount of JVM heap. See attributes/default.rb
for those settings.
elkstack::kibana
Leans on the upstream lusis/chef-kibana
cookbook for most of its work. Sets up
an nginx site for kibana by default. By default, it also does not pass through
most of the http paths directly to elasticsearch (whitelist).
elkstack::java
Wrapper for a java recipe. This is not included on the run list normally, so if
you don't already, you must include this recipe or get another JVM installed
before including anything else in this cookbook.
elkstack::newrelic
Validates if there is a newrelic license set and based on that, see if the node
is tagged as 'elkstack' or 'elkstack_cluster' and creates a file with
elasticsearch details. Installs python, pip and setuptools packages in order to
support newrelic_meetme_plugin
elkstack::acl
Adds cluster node basic iptables rules and cluster iptables rules if appropriate attributes
are set.
elkstack::agent_acl
Adds agent node basic iptables rules if appropriate attributes are set.
elkstack::disk_setup
Look for node['elkstack']['config']['data_disk']['disk_config_type']
to be truthy, and configure the upstream elasticsearch cookbook to format, mount, and use devices appropriately.
elkstack::*_monitoring
These correspond with the recipes above, and just provide a way to pull out the
monitoring work to make the original recipes cleaner.
Miscellaneous
The wrapper recipes are single
and cluster
. These change attributes and then
invoke elasticsearch
, logstash
, kibana
, and rsyslog
. Finally, there are
utility recipes like java
and newrelic
(not invoked otherwise), as well as
acl
which is called by _base
if node['elkstack']['iptables']['enabled']
.
Contributing
See CONTRIBUTING.
Authors
Author:: Rackspace (devops-chef@rackspace.com)
License
# Copyright 2014, Rackspace Hosting
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
Dependent cookbooks
Contingent cookbooks
2.1.3
- Add attribute for additional templates to populate when this cookbook configures a logstash agent
- Correct platformstack/logging-enabled check to be more accurate
- Move lumberjack back to using JSON for exchanging data
2.1.2
- Fix an upstream logstash cookbook issue, contribute it back upstream (#360). Once that is merged, we can go back to upstream.
- Move agent attributes out more, to be their own explicit settings.
2.1.1
- Split out agent attributes into new attribute file.
- Clamp down agent memory usage from 256M to 10% of system at most.
- Fix logstash version typo, 1.4.1 vs. 1.4.2.
2.1.0
- Bump elasticsearch to version 1.3.3. This should improve memory consumption and has a host of other bug fixes. See release notes.
2.0.1
- Add log warning for when we are automatically generating lumberjack keypairs
- Updates to README about memory consumption and how to optimize it
- Moved variables around in agent to make intent more clear (no logic changes)
2.0.0
- Add a
forwarder.rb
recipe that installs logstash-forwarder as an alternative to logstash as an agent, including unit and integration tests. - Add additional tests for existing test-kitchen suites to ensure new lumberjack keypair is written to disk.
- Fix a chefspec test issue where tests were checking for something that didn't make sense, didn't pass.
1.1.0
- Add
chef_environment
fields to be shipped to central cluster - Add ACL recipe for agents to open up to :9300 on the ELK cluster nodes
- Fix protocol from node to transport on logstash agents (to avoid needing more ACLs)
1.0.0
- Bump logstash to version 1.4.2.
- Added top-level
agent
recipe and test suite, intended for running logstash-agent on a regular server. Useselasticsearch::search_discovery
to find and list elk cluster nodes (can override this as well, see elasticsearch cookbook for how). Agent uses lumberjack protocol by default and requires a keypair before converging. - Switch nodes to communicate using node protocol on
eslocal:9300
between logstash and elasticsearch on the same box. - Write chefspec tests for default (server nodes) and agent to 100% coverage. Add
chef-sugar
toGemfile
to be able to converge with chefspec and fauxhai. - Drop dependency on
logstash_stack
(didn't need to be there any longer, wasn't used). - Nodes no longer forward directly to logstash on the remote side, they forward to a local logstash listening for syslog.
- Bugfix:
/etc/hosts
is now correctly populated for all classes of elkstack, not just multi-node. - Bugfix: The
newrelic
plugin now monitors against eslocal, not localhost.
0.3.0
- Bumped default Elasticsearch version to 1.3.2. Java 7 is now required, and some tests had to be adjusted.
- Fix bug in username not being used in tests because it was wrong in /etc/nginx/htpassword.curl.
- Fix bug where port 443 was not open in iptables.
0.2.0
- Added the ability to disable redirects on kibana
0.1.3
- Sheppy Reno - Convert process monitors to platformstack
0.1.2
- Add more options for kibana username and password fields under basic auth over SSL on nginx.
0.1.1
- Seperate recipes per service, add searching and tests. Major workarounds for logstash cookbook.
0.1.0
- Initial release of elkstack
Foodcritic Metric
2.1.4 passed this metric
2.1.4 passed this metric