- DCOS_OSS-5970 Schedules can be defined with job spec creation. They are always returned.
èmbed=schedules
has been deprecated.
- Provide configuration in
/info
. - D2IQ-69445 Introduce
/leader
endpoint to query the current Metronome leader address in HA mode.
-
DCOS_OSS-5965 Fix an issue Metronome became unresponsive when too many
JobRunExecutorActor
instances queried instances and timed out. -
MARATHON-8746 Fix a severe issue introduced in 0.6.41 in which the zk-store location was modified, causing Metronome to appear that it lost all jobs. Upgrading to 0.6.48 will resolves this issue and restores jobs created in previous versions of DC/OS.
- Allow Metronome jobs to specify networks to join MARATHON-8727
- Optimize JsonSchema validation for JobSpec ids MARATHON-8730
- Fixes issue in which JobSpec history sometimes wasn't properly cleaned up DCOS_OSS-3997
- Fixes an issue with DST DCOS_OSS-2509
- Updated to the latest version of cron-utils
9.0.0
and removed threeten-backport. This fixes a number of cron related issues in the underlying dependencies. - When querying run detail with
embed=history
,successfulFinishedRuns
andfailedFinishedRuns
contains new fieldtasks
which is an array of taskIds of that finished run. This will allow people to query task ids even for finished job runs. - Fixed a bug when task status was not updated after the task turned running (when querying embed=activeRuns).
- DCOS_OSS-5166 Fixed metronome not using revive operation
- DCOS_OSS-5020 Add missing HTTP metrics in Metronome.
- DCOS_OSS-4636 Failure when restart policy is
ON_FAILURE
. This bug was introduced through the fix of another bug regarding stopping invalid extra instances of a job run. Metronome should not check the launch queue when a restart is invoked.
- Added new metric
metronome.uptime.gauge.seconds
- DCOS_OSS-4978 Allow using
IS
operator when creating jobs. This was broken since introduction of theIS
operator, which replacedEQ
but was not a valid schema value.
Updated Marathon dependency to 1.7.202
Updated Marathon dependency to 1.7.188
Metronome uses Marathon as a library for scheduling. We have bumped the dependency to the current Marathon, which is 1.7.183. This brings a lot of bug fixes and new features from the last 3 versions of Marathon. At the same time, it allows us to add UCR and secrets support.
Metronome 0.5.71 contains new Metrics endpoint with new metrics exposed that should allow you to monitor Metronome more easily. For detailed information please refer to the Metrics page in our docs.
In order to bring better alignment between Marathon and Metronome, the EQ
constraint operator has been replaced with IS
. The change is semantic; Job definitions using EQ
will continue to function the same and are transparently mapped to the new operator with the same constraint behavior.
If you post the following Job definition:
{
"description": "constraint example",
"id": "constraint-example",
"run": {
...
"placement": {
"constraints": [{"attribute": "@region", "operator": "EQ", "value": "us-east-1"}]
}
}
}
When you ask for it back, the operator will be "IS":
{
"description": "constraint example",
"id": "constraint-example",
"run": {
...
"placement": {
"constraints": [{"attribute": "@region", "operator": "IS", "value": "us-east-1"}]
}
}
}
Previous jobs are automatically migrated as well.
- DCOS_OSS-4344 Support UCR
- DCOS_OSS-4464 EQ operator is replaced with IS (in backward compatible way)
- DCOS_OSS-4446 Support file based secrets
- DCOS_OSS-4440 GPU support
- DCOS_OSS-4024 Use newer Caffeine dependency
- DCOS_OSS-4239 Crash when Zookeeper connection fails to establish
- #244 Wait for all parts of migration to be finished.
Diff 0.4.2-0.4.3
- #234 Exit when cannot load state from ZK.
- #230 Gracefully handle errors during task launching.
- DCOS_OSS-2564 Docker params support.
Diff 0.4.2-0.4.3
- METRONOME-248 Environment variable secrets exposed via API
- METRONOME-218 Improved behavior for situations when the underlying zookeeper node content is corrupt - we now fail loud and early
Diff 0.4.1-0.4.2
Command line parameter task.lost.expunge.gc
was removed because the underlying algorithm change and this
one no longer has any effect.
We moved to a different Metrics library and the metrics are not always compatible or the same as existing metrics; however, the metrics are also now more accurate, use less memory, and are expected to get better throughout the release. Where it was possible, we maintained the original metric names/groupings/etc, but some are in new locations or have slightly different semantics. Any monitoring dashboards should be updated.
For Metronome specific metrics, you can find your old metrics under in the same path, only prefixed with "service" so e.g. 'dcos.metronome.jobspec.impl.JobSpecServiceActor.receiveTimer' is now 'service.dcos.metronome.jobspec.impl.JobSpecServiceActor.receiveTimer'.
The format of the v1/metrics endpoint also changed in a backward incompatible manner - please see the documentation for the current way the metrics are served.
- METRONOME-222 CMD or Docker is Required.
- METRONOME-236 Additional CRON validation to prevent system lock up.
Diff 0.4.0-0.4.1
- METRONOME-190 Added launch queue
- METRONOME-194 Support FORBID Concurrency Policy
- METRONOME-100 Metronome restart causes duplication of jobrun
- METRONOME-191 Implement startingDeadlineTimeout
Diff 0.3.4-0.4.0
The launch queue (/v1/queue
) provides a way to see jobs which have been scheduled to launch but are still not launched on the cluster.
This is usually because there is not enough resources or constraints are not met.
The FORBID concurrency policy allows "concurrencyPolicy": "FORBID"
to be added to a schedule. This restricts launching of a scheduled jobrun when previous run is still active. In that case it will not launch nor will be queued to launch. The job will be rescheduled for the next CRON time.
Diff 0.3.4-0.3.5
- METRONOME-236 Additional CRON validation to prevent system lock up.
Diff 0.3.3-0.3.4
- METRONOME-207 V0 Endpoint needs to support ForcePullImage
Diff 0.3.2-0.3.3
- METRONOME-188 Updated to Protocol Buffers v.3.3.0
- METRONOME-196 ForcePullImage should not be required
Diff 0.2.4-0.3.1
- Upgraded to a released version of Marathon Lib v1.3.13
- Updates to dependencies (including an Akka update to fix schedule time wrap around bug)
- Added
/info
end point for metronome version information
- #150 Added
/info
endpoint - MARATHON_EE-1717 60s min between reschedules
- MARATHON_EE-1725
- MARATHON_EE-1726 Upgrade Marathon libraries and dependencies
https://s3.amazonaws.com/downloads.mesosphere.io/metronome/releases/0.3.1/metronome-0.3.1.tgz sha: f6fd3d48a889ea19cb13dfd908a82e53c03ffab1
diff from 0.2.3-0.2.4
- Upgraded to cron-utils 6.0.4, fixes issues with cron calculations enabling crons such as
0 9 1-7 * 1-5
as mon-fri the first week of the month only. - Documentation and Job placement examples provided.
Prepare Metronome for DC/OS 1.9.
We now have a suite of integration tests for DC/OS that runs in our CI.
- Fix #96 Change constraints fields names to match schema.
- Fix #102 API Examples update: test.com -> example.com
- Fix #107 Add /v0/scheduled-jobs raml documentation.
Genesis