Ukraine flag We stand with our friends and colleagues in Ukraine. To support Ukraine in their time of need visit this page.

Deployment


See also:


The main Jaeger backend components are released as Docker images on Docker Hubexternal link and Quayexternal link :

ComponentDocker HubQuay
jaeger-agenthub.docker.com/r/jaegertracing/jaeger-agent/external linkquay.io/repository/jaegertracing/jaeger-agentexternal link
jaeger-collectorhub.docker.com/r/jaegertracing/jaeger-collector/external linkquay.io/repository/jaegertracing/jaeger-collectorexternal link
jaeger-queryhub.docker.com/r/jaegertracing/jaeger-query/external linkquay.io/repository/jaegertracing/jaeger-queryexternal link
jaeger-ingesterhub.docker.com/r/jaegertracing/jaeger-ingester/external linkquay.io/repository/jaegertracing/jaeger-ingesterexternal link

There are orchestration templates for running Jaeger with:

Configuration Options

Jaeger binaries can be configured in a number of ways (in the order of decreasing priority):

  • command line arguments,
  • environment variables,
  • configuration files in JSON, TOML, YAML, HCL, or Java properties formats.

To see the complete list of options, run the binary with help command or refer to the CLI Flags page for more information. Options that are specific to a certain storage backend are only listed if the storage type is selected. For example, to see all available options in the Collector with Cassandra storage:

$ docker run --rm \
    -e SPAN_STORAGE_TYPE=cassandra \
    jaegertracing/jaeger-collector:1.22 \
    help

In order to provide configuration parameters via environment variables, find the respective command line option and convert its name to UPPER_SNAKE_CASE, for example:

Command line optionEnvironment variable
--cassandra.connections-per-hostCASSANDRA_CONNECTIONS_PER_HOST
--metrics-backendMETRICS_BACKEND

Agent

Jaeger client libraries expect jaeger-agent process to run locally on each host. The agent exposes the following ports:

PortProtocolFunction
6831UDPaccept jaeger.thriftexternal link in compact Thrift protocol used by most current Jaeger clients
6832UDPaccept jaeger.thriftexternal link in binary Thrift protocol used by Node.js Jaeger client (because thriftrwexternal link npm package does not support compact protocol)
5778HTTPserve configs, sampling strategies
5775UDPaccept zipkin.thriftexternal link in compact Thrift protocol (deprecated; only used by very old Jaeger clients, circa 2016)
14271HTTPadmin port: health check at / and metrics at /metrics

It can be executed directly on the host or via Docker, as follows:

## make sure to expose only the ports you use in your deployment scenario!
docker run \
  --rm \
  -p6831:6831/udp \
  -p6832:6832/udp \
  -p5778:5778/tcp \
  -p5775:5775/udp \
  jaegertracing/jaeger-agent:1.22

Discovery System Integration

The agents can connect point to point to a single collector address, which could be load balanced by another infrastructure component (e.g. DNS) across multiple collectors. The agent can also be configured with a static list of collector addresses.

On Docker, a command like the following can be used:

docker run \
  --rm \
  -p5775:5775/udp \
  -p6831:6831/udp \
  -p6832:6832/udp \
  -p5778:5778/tcp \
  jaegertracing/jaeger-agent:1.22 \
  --reporter.grpc.host-port=jaeger-collector.jaeger-infra.svc:14250

When using gRPC, you have several options for load balancing and name resolution:

  • Single connection and no load balancing. This is the default if you specify a single host:port. (example: --reporter.grpc.host-port=jaeger-collector.jaeger-infra.svc:14250)
  • Static list of hostnames and round-robin load balancing. This is what you get with a comma-separated list of addresses. (example: reporter.grpc.host-port=jaeger-collector1:14250,jaeger-collector2:14250,jaeger-collector3:14250)
  • Dynamic DNS resolution and round-robin load balancing. To get this behaviour, prefix the address with dns:/// and gRPC will attempt to resolve the hostname using SRV records (for external load balancingexternal link ), TXT records (for service configsexternal link ), and A records. Refer to the gRPC Name Resolution docsexternal link and the dns_resolver.go implementationexternal link for more info. (example: --reporter.grpc.host-port=dns:///jaeger-collector.jaeger-infra.svc:14250)

Agent level tags

Jaeger supports agent level tags, that can be added to the process tags of all spans passing through the agent. This is supported through the command line flag --agent.tags=key1=value1,key2=value2,...,keyn=valuen. Tags can also be set through an environment flag like so - --agent.tags=key=${envFlag:defaultValue} - The tag value will be set to the value of the envFlag environment key and defaultValue if not set.

Collectors

The collectors are stateless and thus many instances of jaeger-collector can be run in parallel. Collectors require almost no configuration, except for the location of Cassandra cluster, via --cassandra.keyspace and --cassandra.servers options, or the location of Elasticsearch cluster, via --es.server-urls, depending on which storage is specified. To see all command line options run

go run ./cmd/collector/main.go -h

or, if you don’t have the source code

docker run -it --rm jaegertracing/jaeger-collector:1.22 -h

At default settings the collector exposes the following ports:

PortProtocolFunction
14250gRPCused by jaeger-agent to send spans in model.proto format
14268HTTPcan accept spans directly from clients in jaeger.thrift format over binary thrift protocol
9411HTTPcan accept Zipkin spans in Thrift, JSON and Proto (disabled by default)
14269HTTPadmin port: health check at / and metrics at /metrics

Storage Backends

Collectors require a persistent storage backend. Cassandra and Elasticsearch are the primary supported storage backends. Additional backends are discussed hereexternal link .

The storage type can be passed via SPAN_STORAGE_TYPE environment variable. Valid values are cassandra, elasticsearch, kafka (only as a buffer), grpc-plugin, badger (only with all-in-one) and memory (only with all-in-one).

As of version 1.6.0, it’s possible to use multiple storage types at the same time by providing a comma-separated list of valid types to the SPAN_STORAGE_TYPE environment variable. It’s important to note that all listed storage types are used for writing, but only the first type in the list will be used for reading and archiving.

For large scale production deployment the Jaeger team recommends Elasticsearch backend over Cassandra .

Memory

The in-memory storage is not intended for production workloads. It’s intended as a simple solution to get started quickly and data will be lost once the process is gone.

By default, there’s no limit in the amount of traces stored in memory but a limit can be established by passing an integer value via --memory.max-traces.

Badger - local storage

Experimental since Jaeger 1.9

Badgerexternal link is an embedded local storage, only available with all-in-one distribution. By default it acts as an ephemeral storage using a temporary filesystem. This can be overridden by using the --badger.ephemeral=false option.

docker run \
  -e SPAN_STORAGE_TYPE=badger \
  -e BADGER_EPHEMERAL=false \
  -e BADGER_DIRECTORY_VALUE=/badger/data \
  -e BADGER_DIRECTORY_KEY=/badger/key \
  -v <storage_dir_on_host>:/badger \
  -p 16686:16686 \
  jaegertracing/all-in-one:1.22

Cassandra

Supported versions: 3.4+

Deploying Cassandra itself is out of scope for our documentation. One good source of documentation is the Apache Cassandra Docsexternal link .

Configuration

Minimal
docker run \
  -e SPAN_STORAGE_TYPE=cassandra \
  -e CASSANDRA_SERVERS=<...> \
  jaegertracing/jaeger-collector:1.22
All options

To view the full list of configuration options, you can run the following command:

docker run \
  -e SPAN_STORAGE_TYPE=cassandra  \
  jaegertracing/jaeger-collector:1.22 \
  --help

Schema script

A script is provided to initialize Cassandra keyspace and schema using Cassandra’s interactive shell cqlshexternal link :

MODE=test sh ./plugin/storage/cassandra/schema/create.sh | cqlsh

For production deployment, pass MODE=prod DATACENTER={datacenter} arguments to the script, where {datacenter} is the name used in the Cassandra configuration / network topology.

The script also allows overriding TTL, keyspace name, replication factor, etc. Run the script without arguments to see the full list of recognized parameters.

TLS support

Jaeger supports TLS client to node connections as long as you’ve configured your Cassandra cluster correctly. After verifying with e.g. cqlsh, you can configure the collector and query like so:

docker run \
  -e CASSANDRA_SERVERS=<...> \
  -e CASSANDRA_TLS=true \
  -e CASSANDRA_TLS_SERVER_NAME="CN-in-certificate" \
  -e CASSANDRA_TLS_KEY=<path to client key file> \
  -e CASSANDRA_TLS_CERT=<path to client cert file> \
  -e CASSANDRA_TLS_CA=<path to your CA cert file> \
  jaegertracing/jaeger-collector:1.22

The schema tool also supports TLS. You need to make a custom cqlshrc file like so:

# Creating schema in a cassandra cluster requiring client TLS certificates.
#
# Create a volume for the schema docker container containing four files:
# cqlshrc: this file
# ca-cert: the cert authority for your keys
# client-key: the keyfile for your client
# client-cert: the cert file matching client-key
#
# if there is any sort of DNS mismatch and you want to ignore server validation
# issues, then uncomment validate = false below.
#
# When running the container, map this volume to /root/.cassandra and set the
# environment variable CQLSH_SSL=--ssl
[ssl]
certfile = ~/.cassandra/ca-cert
userkey = ~/.cassandra/client-key
usercert = ~/.cassandra/client-cert
# validate = false

Elasticsearch

Supported in Jaeger since 0.6.0 Supported versions: 5.x, 6.x, 7.x

Elasticsearch version is automatically retrieved from root/ping endpoint. Based on this version Jaeger uses compatible index mappings and Elasticsearch REST API. The version can be explicitly provided via --es.version= flag.

Elasticsearch does not require initialization other than installing and running Elasticsearchexternal link . Once it is running, pass the correct configuration values to the Jaeger collector and query service.

Configuration

Minimal
docker run \
  -e SPAN_STORAGE_TYPE=elasticsearch \
  -e ES_SERVER_URLS=<...> \
  jaegertracing/jaeger-collector:1.22
All options

To view the full list of configuration options, you can run the following command:

docker run \
  -e SPAN_STORAGE_TYPE=elasticsearch \
  jaegertracing/jaeger-collector:1.22 \
  --help

Shards and Replicas for Elasticsearch indices

Shards and replicas are some configuration values to take special attention to, because this is decided upon index creation. This articleexternal link goes into more information about choosing how many shards should be chosen for optimization.

Elasticsearch Rollover

Elasticsearch rolloverexternal link is an index management strategy that optimizes use of resources allocated to indices. For example, indices that do not contain any data still allocate shards, and conversely, a single index might contain significantly more data than the others. Jaeger by default stores data in daily indices which might not optimally utilize resources. Rollover feature can be enabled by --es.use-aliases=true.

Rollover lets you configure when to roll over to a new index based on one or more of the following criteria:

Rollover index management strategy is more complex than using the default daily indices and it requires an initialisation job to prepare the storage and two cron jobs to manage indices.

To learn more about rollover index management in Jaeger refer to this articleexternal link .

For automated rollover, please refer to Elasticsearch ILM support .

Initialize

The following command prepares Elasticsearch for rollover deployment by creating index aliases, indices, and index templates:

docker run -it --rm --net=host jaegertracing/jaeger-es-rollover:latest init http://localhost:9200 # <1>

If you need to initialize archive storage, add -e ARCHIVE=true.

After the initialization Jaeger can be deployed with --es.use-aliases=true.

Rolling over to a new index

The next step is to periodically execute the rollover API which rolls the write alias to a new index based on supplied conditions. The command also adds a new index to the read alias to make new data available for search.

docker run -it --rm --net=host -e CONDITIONS='{"max_age": "2d"}' jaegertracing/jaeger-es-rollover:latest rollover  http://localhost:9200 # <1>

<1> The command rolls the alias over to a new index if the age of the current write index is older than 2 days. For more conditions see Elasticsearch docsexternal link .

The next step is to remove old indices from read aliases. It means that old data will not be available for search. This imitates the behavior of --es.max-span-age flag used in the default index-per-day deployment. This step could be optional and old indices could be simply removed by index cleaner in the next step.

docker run -it --rm --net=host -e UNIT=days -e UNIT_COUNT=7 jaegertracing/jaeger-es-rollover:latest lookback  http://localhost:9200 # <1>

<1> Removes indices older than 7 days from read alias.

Remove old data

The historical data can be removed with the jaeger-es-index-cleaner that is also used for daily indices.

docker run -it --rm --net=host -e ROLLOVER=true jaegertracing/jaeger-es-index-cleaner:latest 14 http://localhost:9200 # <1>

<1> Remove indices older than 14 days.

Elasticsearch ILM support

Elasticsearch ILMexternal link automatically manages indices according to performance, resiliency, and retention requirements.

For example:

  • Rollover to a new index by size (bytes or number of documents) or age, archiving previous indices
  • Delete stale indices to enforce data retention standards
Enabling ILM support
  • Create an ILM policy in elasticsearch named jaeger-ilm-policy.

    For example, the following policy will rollover the “active” index when it is older than 1m and delete indices that are older than 2m.

    curl -X PUT http://localhost:9200/_ilm/policy/jaeger-ilm-policy \
    -H 'Content-Type: application/json; charset=utf-8' \
    --data-binary @- << EOF
    {
      "policy": {
        "phases": {
          "hot": {
            "min_age": "0ms",
            "actions": {
              "rollover": {
                "max_age": "1m"
              },
              "set_priority": {
                "priority": 100
              }
            }
          },
          "delete": {
            "min_age": "2m",
            "actions": {
              "delete": {}
            }
          }
        }
      }
    }
    EOF
    
  • Run elasticsearch initializer with ES_USE_ILM=true:

    docker run -it --rm --net=host -e ES_USE_ILM=true jaegertracing/jaeger-es-rollover:latest init http://localhost:9200 # <1>
    

    <1> If you need to initialize archive storage, add -e ARCHIVE=true.

    While initializing with ILM support, make sure that an ILM policy named jaeger-ilm-policy is created in Elasticsearch beforehand (see the previous step), otherwise the following error message will be shown:

    “ILM policy jaeger-ilm-policy doesn’t exist in Elasticsearch. Please create it and rerun init”

    After the initialization, deploy Jaeger with --es.use-ilm=true, --es.use-aliases=true and --es.create-index-templates=false.

Upgrade Elasticsearch version

Elasticsearch defines wire and index compatibility versions. The index compatibility defines the minimal version a node can read data from. For example Elasticsearch 7 can read indices created by Elasticsearch 6, however it cannot read indices created by Elasticsearch 5 even though they use the same index mappings. Therefore upgrade from Elasticsearch 6 to 7 does not require any data migration. However, upgrade from Elasticsearch 5 to 7 has to be done through Elasticsearch 6 and wait until indices created by ES 5.x are removed or explicitly reindexed.

Refer to the Elasticsearch documentationexternal link for wire and index compatibility versions. Generally this information can be retrieved from root/ping REST endpoint.

Reindex

Manual reindexing can be used when upgrading from Elasticsearch 5 to 7 (through Elasticsearch 6) without waiting until indices created by Elasticsearch 5 are removed.

  1. Reindex all span indices to new indices with suffix -1:

    curl -ivX POST -H "Content-Type: application/json" http://localhost:9200/_reindex -d @reindex.json
    {
      "source": {
        "index": "jaeger-span-*"
      },
      "dest": {
        "index": "jaeger-span"
      },
      "script": {
        "lang": "painless",
        "source": "ctx._index = 'jaeger-span-' + (ctx._index.substring('jaeger-span-'.length(), ctx._index.length())) + '-1'"
      }
    }
    
  2. Delete indices with old mapping:

    curl -ivX DELETE -H "Content-Type: application/json" http://localhost:9200/jaeger-span-\*,-\*-1
    
  3. Create indices without -1 suffix:

    curl -ivX POST -H "Content-Type: application/json" http://localhost:9200/_reindex -d @reindex.json
    {
      "source": {
        "index": "jaeger-span-*"
      },
      "dest": {
        "index": "jaeger-span"
      },
      "script": {
        "lang": "painless",
        "source": "ctx._index = 'jaeger-span-' + (ctx._index.substring('jaeger-span-'.length(), ctx._index.length() - 2))"
      }
    }
    
  4. Remove suffixed indices:

    curl -ivX DELETE -H "Content-Type: application/json" http://localhost:9200/jaeger-span-\*-1
    

Run the commands analogically for other Jaeger indices.

There might exist more effective migration procedure. Please share with the community any findings.

Kafka

Supported in Jaeger since 1.6.0 Supported Kafka versions: 0.9+

Kafka can be used as an intermediary buffer between collector and an actual storage. The collector is configured with SPAN_STORAGE_TYPE=kafka that makes it write all received spans into a Kafka topic. A new component Ingester , added in version 1.7.0, is used to read from Kafka and store spans in another storage backend (Elasticsearch or Cassandra).

Writing to Kafka is particularly useful for building post-processing data pipelines.

Configuration

Minimal
docker run \
  -e SPAN_STORAGE_TYPE=kafka \
  -e KAFKA_PRODUCER_BROKERS=<...> \
  -e KAFKA_TOPIC=<...> \
  jaegertracing/jaeger-collector:1.22
All options

To view the full list of configuration options, you can run the following command:

docker run \
  -e SPAN_STORAGE_TYPE=kafka \
  jaegertracing/jaeger-collector:1.22 \
  --help

Topic & partitions

Unless your Kafka cluster is configured to automatically create topics, you will need to create it ahead of time. You can refer to the Kafka quickstart documentationexternal link to learn how.

You can find more information about topics and partitions in general in the official documentationexternal link . This articleexternal link provide more details about how to choose the number of partitions.

Storage plugin

Jaeger supports gRPC based storage plugins. For more information refer to jaeger/plugin/storage/grpcexternal link

Available plugins:

docker run \
  -e SPAN_STORAGE_TYPE=grpc-plugin \
  -e GRPC_STORAGE_PLUGIN_BINARY=<...> \
  -e GRPC_STORAGE_PLUGIN_CONFIGURATION_FILE=<...> \
  jaegertracing/all-in-one:1.22

Ingester

jaeger-ingester is a service which reads span data from Kafka topic and writes it to another storage backend (Elasticsearch or Cassandra).

PortProtocolFunction
14270HTTPadmin port: health check at / and metrics at /metrics

To view all exposed configuration options run the following command:

docker run \
  -e SPAN_STORAGE_TYPE=cassandra \
  jaegertracing/jaeger-ingester:1.22
  --help

Query Service & UI

jaeger-query serves the API endpoints and a React/Javascript UI. The service is stateless and is typically run behind a load balancer, such as NGINXexternal link .

At default settings the query service exposes the following port(s):

PortProtocolFunction
16685gRPCProtobuf/gRPC QueryServiceexternal link
16686HTTP/api/* endpoints and Jaeger UI at /
16687HTTPadmin port: health check at / and metrics at /metrics

Minimal deployment example (Elasticsearch backend):

docker run -d --rm \
  -p 16685:16685 \
  -p 16686:16686 \
  -p 16687:16687 \
  -e SPAN_STORAGE_TYPE=elasticsearch \
  -e ES_SERVER_URLS=http://<ES_SERVER_IP>:<ES_SERVER_PORT> \
  jaegertracing/jaeger-query:1.22

Clock Skew Adjustment

Jaeger backend combines trace data from applications that are usually running on different hosts. The hardware clocks on the hosts often experience relative drift, known as the clock skew effectexternal link . Clock skew can make it difficult to reason about traces, for example, when a server span may appear to start earlier than the client span, which should not be possible. The query service implements a clock skew adjustment algorithm ( codeexternal link ) to correct for clock drift, using the knowledge about causal relationships between spans. All adjusted spans have a warning displayed in the UI that provides the exact clock skew delta applied to its timestamps.

Sometimes these adjustments themselves make the trace hard to understand. For example, when repositioning the server span within the bounds of its parent span, Jaeger does not know the exact relationship between the request and response latencies, so it assumes then to be equal and places the child span in the middle of the parent span (see issue #961external link ).

The query service supports a configuration flag --query.max-clock-skew-adjustment that controls how much clock skew adjustment should be allowed. Setting this parameter to zero (0s) disables clock skew adjustment completely. This setting applies to all traces retrieved from the given query service. There is an open ticket #197external link to support toggling the adjustment on and off directly in the UI.

UI Base Path

The base path for all jaeger-query HTTP routes can be set to a non-root value, e.g. /jaeger would cause all UI URLs to start with /jaeger. This can be useful when running jaeger-query behind a reverse proxy.

The base path can be configured via the --query.base-path command line parameter or the QUERY_BASE_PATH environment variable.

UI Customization and Embedding

Please refer to the dedicated Frontend/UI page .

Aggregation Jobs for Service Dependencies

Production deployments need an external process which aggregates data and creates dependency links between services. Project spark-dependenciesexternal link is a Spark job which derives dependency links and stores them directly to the storage.