This document constitutes a per-version listing of changes of configuration which are non-backwards compatible.
The script for geo_enrichment_load.sh
has been renamed, and now is maxmind_enrichment_load.sh
. A couple changes should happen for users who are upgrading.
- The MaxMind GeoLite2 ASN database should be loaded onto HDFS at /apps/metron/asn/default/GeoLite2-ASN.tar.gz OR the global configuration property
asn.hdfs.file
can be set to point to a custom HDFS location. - Any custom scripts or tasks that use this script should be updated. In addition, this updated script also retrieves the GeoLite2 ASN database. The
-ra
flag can be used to provide a custom location for this database if offline install is needed. Otherwise, it will retrieve the latest from MaxMind.
The Elasticsearch Java client has now been migrated from TransportClient to the new Java REST client. The motivation for this change
is that TransportClient will be deprecated in Elasticsearch 7.0 and removed entirely in 8.0. See ES Java API for more details.
The primary client-facing change for upgrades will be the new properties for configuring the new client. An explanation of the new properties
as well as a mapping from the old properties to the new can be found in metron-elasticsearch under es.client.settings
.
The unified enrichment topology will be the new default in this release, and the split-join enrichment topology is now considered deprecated. If you wish to keep the deprecated split-join enrichment topology, you will need to make the following changes:
-
In Ambari > Metron > Config > Enrichment set the enrichment_topology setting to "Split-Join"
-
If running
start_enrichment_topology.sh
manually, pass in the parameters to start the Split-Join topology as follows$METRON_HOME/bin/start_enrichment_topology.sh --remote $METRON_HOME/flux/enrichment/remote-splitjoin.yaml --filter $METRON_HOME/config/enrichment-splitjoin.properties
-
Restart the enrichment topology
While modifying the PaloAlto log parser to support logs from newer PAN-OS version and to not break when a message payload contains a comma, some field names were changed to extend the coverage, fix some duplicate names and change some field names to the Metron standard message format.
Installations making use of this parser should check, if the resulting messages still meet their expectations and adjust downstream configurations (i.e. ElasticSearch template) accordingly.
Note: Previously, the samples for the test contained a full syslog line (including syslog header). This did - and will continue to - create a broken "domain" field in the parsed message. It is recommended to only feed the syslog message part to the parser for now.
As we continue to evolve the Stellar language, it is possible that new keywords will be added to the language. This may cause compatablity issues where these reserved words and symbols are used in existing scripts.
Adding match
to the Stellar lanaguage has introduced the following new
reserved keywords and symbols:
match
, default
, {
, }
, '=>'
Any stellar expressions which use these keywords not in quotes will need to be modified.
In order to allow for meta alerts to be queries alongside regular alerts in Elasticsearch 2.x, it is necessary to add an additional field to the templates and mapping for existing sensors.
Two steps must be done for each sensor, but not on each index for each sensor.
First is to update the Elasticsearch template for each sensor, so any new indices have the field:
export ELASTICSEARCH="node1"
export SENSOR="bro"
curl -XGET "http://${ELASTICSEARCH}:9200/_template/${SENSOR}_index*?pretty=true" -o "${SENSOR}.template"
sed -i '2d;$d' ./${SENSOR}.template
sed -i '/"properties" : {/ a\
"alert": { "type": "nested"},' ${SENSOR}.template
curl -XPUT "http://${ELASTICSEARCH}:9200/_template/${SENSOR}_index" -d @${SENSOR}.template
To update existing indexes, update Elasticsearch mappings with the new field for each sensor. Make sure to set the ELASTICSEARCH variable appropriately.
curl -XPUT "http://${ELASTICSEARCH}:9200/${SENSOR}_index*/_mapping/${SENSOR}_doc" -d '
{
"properties" : {
"alert" : {
"type" : "nested"
}
}
}
'
rm ${SENSOR}.template
For a more detailed description, please see metron-platform/metron-elasticsearch/README.md
In the 0.4.2 release,
Since the Ansible Deployment uses the MPack, RPMs must be built prior to deployment. As a result, Docker is required to perform a Quick-Dev, Full-Dev or Ansible deployment. This effectively limits the build environment to Docker supported platforms.
As of 0.3.0 the indexing configuration
- Is held in the enrichment configuration for a sensor
- Has properties which control every writers (i.e. HDFS, solr or elasticsearch).
In the 0.3.1 release, this configuration has been broken out and control for individual writers are separated.
Please see the description of the configurations in the indexing README
Migrate the configurations from each sensor enrichment configuration and create appropriate configurations for indexing.
For instance, if a sensor enrichment config for sensor foo
is in $METRON_HOME/config/zookeeper/enrichments/foo.json
and looks like
{
"index" : "foo",
"batchSize" : 100
}
You would create a file to configure each writer for sensor foo
called $METRON_HOME/config/zookeeper/indexing/foo.json
with the contents
{
"elasticsearch" : {
"index" : "foo",
"batchSize" : 100,
"enabled" : true
},
"hdfs" : {
"index" : "foo",
"batchSize" : 100,
"enabled" : true
}
}
As of 0.3.0, threat triage rules were defined as a simple Map associating a Stellar expression with a score. As of 0.3.1, due to the fact that there may be many threat triage rules, we have made the rules more complex. To help organize these, we have made the threat triage objects in their own right that contain optional name and optional comment fields.
This essentially makes the risk level rules slightly more complex. The format goes from:
"riskLevelRules" : {
"stellar expression" : numeric score
}
to:
"riskLevelRules" : [
{
"name" : "optional name",
"comment" : "optional comment",
"rule" : "stellar expression",
"score" : numeric score
}
]
For every sensor enrichment configuration, you will need to migrate the riskLevelRules
section
to move from a map to a list of risk level rule objects.
As of 0.3.0, a MySQL database was used for storage and retrieval of GeoIP information during enrichment. As of 0.3.1, the MySQL database is removed in favor of using MaxMind's binary GeoIP files and stored on HDFS
After initial setup, this change is transparent and existing enrichment definitions will run as-is.
While new installs will not require any additional steps, in an existing install a script must be run to retrieve and load the initial data.
The shell script geo_enrichment_load.sh
will retrieve MaxMind GeoLite2
data and load data into HDFS, and update the configuration to point to
this data.
In most cases the following usage will grab the data appropriately:
$METRON_HOME/bin/geo_enrichment_load.sh -z <zk_server>:<zk_port>
Additional options, including changing the source file location (which can be a file:// location if the GeoIP data is already downloaded), are available with the -h flag and are also detailed in the metron-data-management README.me file.
One caveat is that this script will NOT update on disk config files. It is recommended to retrieve the configuration using
$METRON_HOME/bin/zk_load_configs.sh -z <zk_server>:<zk_port> -m DUMP
The new config will be geo.hdfs.file
in the global section of the
configuration. Append this key-value into the global.json in the config
directory. A PUSH is unnecessary
During 0.3.1 we decoupled specifying durations for calls to the profiler
into a separate function. The consequence is that existing calls to
PROFILE_GET
will need to migrate.
Existing calls to PROFILE_GET
will need to change from PROFILE_GET('profile', 'entity', duration, 'durationUnits')
to PROFILE_GET('profile', 'entity', PROFILE_FIXED(duration, 'durationUnits'))
/etc/monit.d/enrichment-elasticsearch.monit
was renamed to
/etc/monit.d/indexing-elasticsearch.monit
, however the old file isn't
removed via ansible, which causes the below error during an upgrade:
Starting monit: /etc/monit.d/enrichment-elasticsearch.monit:18: Service name conflict, enrichment already defined '/usr/local/monit/status_enrichment_topology.sh'
When using Ansible to deploy the latest Metron bits to an existing installation, storm-site is not being updated with the new 0.2.1BETA parameter topology.classpath
. Topologies are unable to find the client configs as a result.
Set the topology.classpath
property for storm in Ambari to /etc/hbase/conf:/etc/hadoop/conf