Monitoring [VC 5.6 INT]
Monitoring
To get better insight into the operations of jobs and services you can collect metrics and traces in your favorite monitoring service. Metrics are exposed using JMX and StatsD.
Transcoders on the other hand only expose metrics using StatsD.
StatsD
By default metrics are not sent to a StatsD server. To enable it you have to update the metrics configuration. For example, to have metrics sent to a StatsD server on localhost
listening on UDP port 8125, use:
PUT API/configuration/metrics
Content-Type: application/xml
<MetricsConfigurationDocument xmlns="http://xml.vidispine.com/schema/vidispine">
<statsd/>
</MetricsConfigurationDocument>
Metrics sent to StatsD are by default prefixed with vs.
. To have metrics sent with the prefix vs1.
, for example if you have multiple instances running:
PUT API/configuration/metrics
Content-Type: application/xml
<MetricsConfigurationDocument xmlns="http://xml.vidispine.com/schema/vidispine">
<statsd>
<host>metrics.example.com</host>
<port>6125</port>
<prefix>vs1</prefix>
</statsd>
</MetricsConfigurationDocument>
Here metrics are sent to an external StatsD server on the non-standard port 6125. Note that the .
between the prefix and metric name is added automatically.
Filtering metrics
You can set inclusion and exclusion filters to restrict which metrics are sent to the StatsD server. The default is to include all and exclude none.
Inclusion/exclusion filters may have a leading or trailing wildcard. For example, to exclude all storage.fs
metrics:
<MetricsConfigurationDocument xmlns="http://xml.vidispine.com/schema/vidispine">
<statsd>
<exclude>storage.fs.*</exclude>
</statsd>
</MetricsConfigurationDocument>
Tagged metrics
Some metrics are tagged with additional information. These are sent to StatsD in the format:
<metricname>:<value>|<type>|#<tag>+
A job.step.execution.time
metric might for example be sent as:
vs.job.step.execution.time:123|ms|#type:placeholder-import,step:100,sync
If your StatsD server does not support such tags then they can be disabled by setting tags
to false:
<MetricsConfigurationDocument xmlns="http://xml.vidispine.com/schema/vidispine">
<statsd>
...
<tags>false</tags>
</statsd>
</MetricsConfigurationDocument>
JMX
Each metric is exposed as an JMX MBean in the “metrics” domain. You can view the metrics using for example:
A JMX client such as VisualVM with the VisualVM-MBeans plugin, or JConsole.
Programmatically using the Java JMX client interface.
Over HTTP/JSON using a bridge such as Jolokia.
Metrics
Metrics are exposed as either meters, timers or gauges. The name of a metric is meant to be self-explanatory. Timers are suffixed with time
and meters are named as past tense verbs, while gauges make up the rest.
The StatsD type used for each metric, and the statistics exposed over JXM for each type are:
Type | StatsD type | MBean attributes |
---|---|---|
Meter |
| The count, mean and 1/5/15-minute rates. |
Gauge |
| The value. |
Timer |
| The count, min/max/mean/stdev, rates and percentiles. |
Indexing
Meters:
reindex.{index}.started
reindex.{index}.finished
indexer.solr.request.failed
indexer.elasticsearch.request.failed
Timers:
indexer.solr.update.time
indexer.solr.delete.time
indexer.solr.commit.time
indexer.elasticsearch.update.time
indexer.elasticsearch.delete.time
indexer.{index}.index.time
With
index
being one ofitem
/collection
/acl
/file
.
indexer.library.update.time
Time spend on updating auto-refreshing libraries in the system.
Job
Meters:
job.created
job.started
job.finished
job.failed
job.blocked
Gauges:
job.total.{state}
Where
state
is the name of a job state, lower cased and with_
replaced with-
. For examplefinished-warning
.
Timers:
job.{type}.step.{step}.{sync}.execution.time
job.step.execution.time
Tagged with
type:{type}
,step:{step}
andsync/async
.
Solr
Meters:
solr.request.failed
Timers:
solr.query.time
solr.update.time
solr.commit.soft.time
solr.commit.hard.time
solr.optimize.time
Elasticsearch
Meters:
elasticsearch.request.failed
Timers:
elasticsearch.query.time
elasticsearch.update.time
elasticsearch.delete.time
Storage
Meters:
storage.online
Tagged with
storage:{id}
.
storage.offline
Tagged with
storage:{id}
.
storage.method.online
Tagged with
storage:{id}
.
storage.method.offline
Tagged with
storage:{id}
.
storage.file.found
Tagged with
storage:{id}
.
storage.file.changed
Tagged with
storage:{id}
.
storage.file.deleted
Tagged with
storage:{id}
.
storage.file.hashed
storage.file.checksum.bytes.read
storage.fs.stat
The number of
stat
call made.
Gauges:
storage.total.online
storage.total.offline
storage.total.evacuating
storage.total.evacuated
The total number of storages with a specific state.
Resource
Meters:
resource.{type}.online
Tagged with
resource:{id}
.
resource.{type}.offline
Tagged with
resource:{id}
.
Agent
Gauges:
agent.total.online
agent.total.offline
The total number of agents with a specific state.
Transfer
Meters:
transfer.bytes.transferred
transfer.started
transfer.finished
transfer.finished-part
transfer.failed
transfer.blocked
Service
Meters:
service.exception
Gauges:
service.load.5
The 5 minute load.
service.load.60
The 60 minute load.
Transcoder
Gauges
transcoder.{transcoder-id}.jobs.running
transcoder.{transcoder-id}.jobs.finished
transcoder.{transcoder-id}.jobs.failed
transcoder.{transcoder-id}.jobs.{transcoder-job-type}.running
transcoder.{transcoder-id}.jobs.{transcoder-job-type}.finished
transcoder.{transcoder-id}.jobs.{transcoder-job-type}.failed
Counters
transcoder.{transcoder-id}.muxer.video.frames
transcoder.{transcoder-id}.encoder.{codec}.frames
transcoder.{transcoder-id}.decoder.{codec}.frames
transcoder.{transcoder-id}.io.{protocol}.{direction}.bytes
Broker
Gauges
broker.queue.{queue}.size
The size of a specific queue. Note that this metric is only present when using the embedded broker.
Cluster
Gauges
cluster.size
The number of members in the cluster.
APM
Vidispine supports application performance monitoring using Elastic APM. It monitors the execution of the application for easy pinpointing of performance issues.
Setup
In order to use Elastic APM you first need to set up an APM server. The elastic APM integration is disabled by default but can be enabled by adding the following configuration to the server.yaml
file:
apm:
elastic:
urls: ["https://localhost:1234/"]
secretToken: secret
serviceName: vidispine
serviceVersion: 5.0
environment: staging
sampleRate: 1
Note
The server will need to restart for any changes to take effect.
Please see the APM configuration reference for details.
Each trace encapsulates an event and may have one of the following types:
request
A HTTP request, either incoming or outgoing.messaging
A JMS message, either incoming or outgoing.scheduled
A single iteration of a scheduled worker.service
A cross-object method invocation of a service layer class.DB
A database query