New detailed documentation for Mesos available via the Mesosphere Website
Chronos is Airbnb's replacement for cron.
It is a distributed and fault-tolerant scheduler that runs on top of Apache Mesos.
You can use it to orchestrate jobs. It supports custom Mesos executors as well
as the default command executor. Thus by default, Chronos executes sh
(on most systems bash) scripts. Chronos can be used to interact with systems
such as Hadoop (incl. EMR), even if the Mesos slaves on which execution happens
do not have Hadoop installed. Included wrapper scripts allow transfering files
and executing them on a remote machine in the background and using asynchronous
callbacks to notify Chronos of job completion or failures.
Chronos has a number of advantages over regular cron. It allows you to schedule your jobs using ISO8601 repeating interval notation, which enables more flexibility in job scheduling. Chronos also supports the definition of jobs triggered by the completion of other jobs. It supports arbitrarily long dependency chains.
Try out the interactive and personalized tutorial for Chronos.
For questions and discussions around Chronos, please use the Google Group "chronos-scheduler": Chronos Scheduler Group. Also join us on IRC in #mesos on freenode.
- Features
- Running Chronos
- Configuring Chronos
- License
- Contributors
- Video Introduction
- Chronos UI
- API
- Job Management
- Debugging Chronos Jobs
- Notes
- Reporting Bugs
- Appendix
If you get an error while compiling Mesos, please consult the FAQ.
- Web UI
- 8601 Repeating Interval Notation
- Handles dependencies
- Job Stats (e.g. 50th, 75th, 95th and 99th percentile timing, failure/success)
- Fault Tolerance (Hot Master)
- Configurable Retries
- Multiple Workers (i.e. Mesos Slaves)
We've included some example run scripts, but the basic syntax for launching chronos is:
java -cp chronos.jar --master zk://127.0.0.1:2181/mesos --zk_hosts 127.0.0.1:2181
Please note that you need to have both Mesos and Zookeeper running for this to work!
For more information on configuration options, please see configuring Chronos.
-
Example runit run script
-
Example local run script
For information on configuring chronos, please see docs/CONFIG.md.
The use and distribution terms for this software are covered by the Apache 2.0 License (http://www.apache.org/licenses/LICENSE-2.0.html) which can be found in the file LICENSE at the root of this distribution. By using this software in any fashion, you are agreeing to be bound by the terms of this license. You must not remove this notice, or any other, from this software.
- Replacing Cron & Building Scalable Data Pipelines: YouTube
Chronos comes with a UI which can be used to add, delete, list, modify and run jobs. It can also show a graph of job dependencies. The screenshot should give you a good idea of what Chronos can do.
You can communicate with Chronos using a RESTful JSON API over HTTP.
Chronos nodes usually listen on port 8080 for API requests.
All examples in this section assume that you've found a running leader at chronos-node.airbnb.com:8080.
When you have multiple Chronos nodes running, only one of them will be elected as the leader. The leader is the only node that responds to API requests, but if you attempt to talk to a non-leader your request will automatically be redirected to a leader.
- Endpoint: /scheduler/jobs
- Method: GET
- Example:
curl -L -X GET chronos-node:8080/scheduler/jobs - Response: JSON data
A job listing returns a JSON list containing all of the jobs. Each job is a JSON hash. Interesting fields in the hashes are:
invocationCount: the number of times the job completedexecutor: auto-determined by Chronos, but will usually be "" for non-async jobsparents: for dependent jobs, a list of all other jobs that must run before this job will do so
If there is a parents field there will be no schedule field and vice-versa.
Get a job name from the job listing above. Then:
- Endpoint: /scheduler/job/jobName
- Method: DELETE
- Example:
curl -L -X DELETE chronos-node:8080/scheduler/job/request_event_counter_hourly - Response: HTTP 204
Deleting tasks for a job is useful if a job gets stuck. Get a job name from the job listing above. Then:
- Endpoint: /scheduler/task/kill/jobName
- Method: DELETE
- Example:
curl -L -X DELETE chronos-node:8080/scheduler/task/kill/request_event_counter_hourly - Response: HTTP 204
You can manually start a job by issuing an HTTP request.
- Endpoint: /scheduler/job
- Method: PUT
- Example:
curl -L -X PUT chronos-node:8080/scheduler/job/request_event_counter_hourly - Response: HTTP 204
The heart of job scheduling is a JSON POST request. The JSON hash you send to Chronos should contain the following fields:
- Name: the job name
- Command: the actual command that will be executed by Chronos
- Schedule: The scheduling for the job, in ISO8601 format. Consists of 3 parts separated by '/':
- Number of times to repeat the job; put just 'R' to repeat forever
- The start time of the job, an empty start time means start immediately
- The run interval
- Epsilon: If Chronos misses the scheduled run time for any reason, it will still run the job if the time is within this interval. Epsilon must be formatted like an ISO 8601 Duration.
- Owner: the email address of the person responsible for the job
- Async: whether the job runs in the background
A job without schedule field will has a "R1//PT24H" schedule by default.
Here is an example job hash:
{
"schedule": "R10/2012-10-01T05:52:00Z/PT2S",
"name": "SAMPLE_JOB1",
"epsilon": "PT15M",
"command": "echo 'FOO' >> /tmp/JOB1_OUT",
"owner": "[email protected]",
"async": false
}Once you've generated the hash, send it to Chronos like so:
-
Endpoint: /scheduler/iso8601
-
Method: POST
-
Example:
curl -L -H 'Content-Type: application/json' -X POST -d '{json hash}' chronos-node:8080/scheduler/iso8601 -
Response: HTTP 204
A dependent job takes the same JSON format as a scheduled job.
However, instead of the schedule field, it will accept a parents field.
This should be a JSON list of all jobs which must run at least once before this job will run.
-
Endpoint: /scheduler/dependency
-
Method: POST
-
Example:
curl -L -X POST -H 'Content-Type: application/json' -d '{dependent hash}' chronos-node:8080/scheduler/dependency
Here is a more elaborate example for a dependency job hash:
{
"async": true,
"command": "bash -x /srv/data-infra/jobs/hive_query.bash run_hive hostings-earnings-summary",
"epsilon": "PT30M",
"errorCount": 0,
"lastError": "",
"lastSuccess": "2013-03-15T13:02:14.243Z",
"name": "hostings_earnings_summary",
"owner": "[email protected]",
"parents": [
"db_export-airbed_hostings",
"db_export-airbed_reservation2s"
],
"retries": 2,
"successCount": 100
}###Adding a Docker Job
A docker job takes the same format as a scheduled job or a dependency job and runs on a docker container. To configure it, an additional container argument is required, which contains a type (req), an image (req), and volumes (optional).
-
Endpoint: /scheduler/iso8601 or /scheduler/depdency
-
Method: POST
-
Example:
curl -L -H 'Content-Type: application/json' -X POST -d '{json hash}' chronos-node:8080/scheduler/iso8601
{
"schedule": "R\/2014-09-25T17:22:00Z\/PT2M",
"name": "dockerjob",
"container": {
"type": "DOCKER",
"image": "libmesos/ubuntu"
},
"cpus": "0.5",
"mem": "512",
"uris": [],
"command": "while sleep 10; do date =u %T; done"
}Chronos allows to describe the dependency graph and has an endpoint to return this graph in form of a dotfile.
- Endpoint: /scheduler/graph/dot
- Method: GET
- Example:
curl -L -X GET chronos-node:8080/scheduler/graph/dotIf your job is long-running, you may want to run it asynchronously. In this case, you need to do two things:
- When adding your job, ensure it is set as asynchronous.
- Your, job, when complete, should reports its completion status to Chronos.
If you forget to do (2), your job will never run again because Chronos will think that it is still running. Reporting job completion to Chronos is done via another API call:
- Endpoint: /scheduler/task/task id
- Method: PUT
- Example:
curl -L -X PUT -H "Content-Type: application/json" -d '{"statusCode":0}' chronos-node:8080/scheduler/task/my_job_run_555_882083xkj302The task id is auto-generated by Chronos. It will be available in your job's environment as $mesos_task_id.
Note: You will probably need to url-encode the mesos task id in order to submit it as part of the URL.
When specifying the command field in your job hash, use the url-runner.bash (make sure it's deployed on all slaves). Alternatively,
you can also use a url in the command field, if your mesos was compiled with cURL libraries.
| Field | Description | Default |
|---|---|---|
| name | Name of job. | - |
| command | Command to execute. | - |
| epsilon | If, for any reason, a job can't be started at the scheduled time, this is the window in which Chronos will attempt to run the job again | PT60S or --task_epsilon. |
| executor | Mesos executor. By default Chronos uses the Mesos command executor. | - |
| executorFlags | Flags to pass to Mesos executor. | - |
| retries | Number of retries to attempt if a command returns a non-zero status | 2 |
| owner | Email addresses to send job failure notifications. Use comma-separated list for multiple addresses. | - |
| async | Execute using Async executor. | false |
| successCount | Number of successes since the job was last modified. | - |
| errorCount | Number of errors since the job was last modified. | - |
| lastSuccess | Date of last successful attempt. | - |
| lastError | Date of last failed attempt. | - |
| cpus | Amount of Mesos CPUs for this job. | 0.1 or --mesos_task_cpu |
| mem | Amount of Mesos Memory in MB for this job. | 128 or --mesos_task_mem |
| disk | Amount of Mesos disk in MB for this job. | 256 or --mesos_task_disk |
| disabled | If set to true, this job will not be run. | false |
| uris | An array of URIs which Mesos will download when the task is started. | - |
| schedule | ISO8601 repeating schedule for this job. If specified, parents must not be specified. |
- |
| parents | An array of parent jobs for a dependent job. If specified, schedule must not be specified. |
- |
| runAsUser | Mesos will run the job as this user, if specified. | --user |
| container | This contains the subfields for the container, type (req), image (req), and volumes (optional). | - |
{
"name":"camus_kafka2hdfs",
"command":"/srv/data-infra/kafka/camus/kafka_hdfs_job.bash",
"epsilon":"PT30M",
"executor":"",
"executorFlags":"",
"retries":2,
"owner":"[email protected]",
"async":false,
"successCount":190,
"errorCount":3,
"lastSuccess":"2014-03-08T16:57:17.507Z",
"lastError":"2014-03-01T00:10:15.957Z",
"cpus":1.0,
"disk":10240,
"mem":1024,
"disabled":false,
"uris":[
],
"schedule":"R/2014-03-08T20:00:00.000Z/PT2H"
}For larger installations, the web UI may be insufficient for managing jobs. At
Airbnb, there are well over 700 production Chronos jobs. Rather than using the
web UI for making edits, we created a script called chronos-sync.rb which can
be used to synchronize configuration from disk to Chronos. For example, you
may have a Git repository that contains all of the Chronos job configurations,
and then you could run an hourly Chronos job that checks out the repository and
runs chronos-sync.rb.
You can initialize the configuration data by running:
$ bin/chronos-sync.rb -u http://chronos/ -p /path/to/jobs/config -c
After that, you can run the normal sync like this:
$ bin/chronos-sync.rb -u http://chronos/ -p /path/to/jobs/config
You can also forcefully update the configuration in Chronos from disk by
passing the -f or --force parameter. In the example above,
/path/to/jobs/config is the path where you would like the configuration data
to live.
Note: chronos-sync.rb does not delete jobs. If you want to remove a job, you must manually remove it using the API or web UI.
Chronos itself can be configured just like dropwizard-logging via the configuration file. If there's something going wrong with the framework itself look here for information. Individual jobs log with their task id on the mesos slaves. Look in the standard out log for your job name and the string "ready for launch", or else "job ct:" and your job name. The job is done when the line in the log says:
Task with id 'value: TASK_ID **FINISHED**
To find debug logs on the mesos slave, look in /tmp/mesos/slaves on the slave instance (unless you've specifically supplied a different log folder for mesos). For example:
/tmp/mesos/slaves/
In that dir, the current slave run is timestamped so look for the most recent. Under that is a list of frameworks; you're interested in the Chronos framework. For example:
/tmp/mesos/slaves/STAMP/frameworks/
The curl executor is even more powerful if the specified URLs are packaged and self-contained executables. This can be done for example via arx, which bundles code into an executable archive. Arx applications in turn contain shell commands and an archive (e.g. a jar file and a startup-script). It's easy to use and there are no libraries required to unpack and execute the archive.
Signed URLs can be used to publish arx files (e.g. on s3).
To start a new scheduler you have to give the JVM access to the native mesos library.
You can do so by either setting the java.library.path to the build mesos library or create an environment variable MESOS_NATIVE_LIBRARY and set it to the mesoslib.dylib / mesoslib.so file
MESOS_NATIVE_LIBRARY: Absolute path to the native mesos library. This is usually/usr/local/lib/libmesos.soon Linux and/usr/local/lib/libmesos.dylibon OSX.MESOS_LAUNCHER_DIR: Absolute path to the src subdirectory of your mesos build, such that the shell executor can be found (e.g. If mesos was built in/Users/florian/airbnb_code/mesos/buildthen the value for this variable would be/Users/florian/airbnb_code/mesos/build/src).MESOS_KILLTREE: Absolute path to the location of thekilltree.shscript. (e.g./Users/florian/airbnb_code/mesos/src/scripts/killtree.sh)
If you're using the installer script this should be setup for you.
To make all of our lives easier we ask that all bug reports include at least the following information:
The output of:
mvn -X clean package
and
java -version
If the error is in running tests, then please include the output of running all the tests.
# Mac/FreeBSD
tail +1 target/surefire-reports/*.txt
# GNU Coreutils
tail -n +1 target/surefire-reports/*.txt
If the error is in the installer, please include all the output from running it with debug enabled:
bash -x bin/installer.bash
If the bug is in building Mesos from scratch, please submit those bugs directly to mesos.
If the bug occurs while running Chronos, please include the following information:
-
The command used to launch Chronos, for example:
java -cp target/chronos.jar com.airbnb.scheduler.Main <args> -
The version of Mesos you are running.
-
The output of
java -version
As we mentioned, Chronos is designed (not required) to run with multiple nodes of which one is elected master.
If you use the cURL command line tool, you can use the -L flag and hit any Chronos node and you will get a
307 REDIRECT to the leader.
Chronos registers itself with Zookeeper at the location /chronos/state. This value can be changed via the configuration file.
Follow these steps to install Chronos on Amazon Linux:
sudo apt-get install autoconf make gcc cpp patch python-dev git libtool default-jdk default-jdk-builddep default-jre gzip libghc-zlib-dev libcurl4-openssl-dev
sudo yum install autoconf make gcc gcc-c++ patch python-devel git libtool java-1.7.0-openjdk-devel zlib-devel libcurl-devel openssl-devel cyrus-sasl-devel
Make sure you're using Java 7: sudo alternatives --config java
git clone https://github.com/apache/mesos.git
cd mesos/
git checkout
export JAVA_HOME=/usr/lib/jvm/java-1.7.0-openjdk-1.7.0.45.x86_64/
./bootstrap
./configure --with-webui --with-included-zookeeper --disable-perftools --enable-frame-pointers
make
sudo make install
Install Node first. On OSX, try brew install node.
Start up Zookeeper, Mesos master, and Mesos slave(s). Then try
export MESOS_NATIVE_LIBRARY=/usr/local/lib/libmesos.so
git clone https://github.com/airbnb/chronos.git
cd chronos
mvn package
java -cp target/chronos*.jar com.airbnb.scheduler.Main --master zk://localhost:5050/mesos --zk_hosts localhost:2181