From version 5.0 onward, Watcher is part of X-Pack. For more information, see
Alerting on cluster and index events.
Release Notes
editRelease Notes
editVersion Compatibility
editYou must run the version of Watcher that matches the version of Elasticsearch you are running. For example, Watcher 2.3.5 requires Elasticsearch 2.3.5.
Change List
edit2.3.5
editAugust 3, 2016
Bug Fixes
- The watch history was not written, if one of the chained inputs resulted in a failure as well as an input containing a field name with dots
- Status of an acked watch, whose condition evaluates to false again is now properly persisted and not lost in case of a master node switch
-
Fixed the watch history template so payloads and request bodies are handled
correctly. To update an existing installation, delete the existing watch history
template by running
DELETE /_template/watch_history
and the correct template will be autocreated. To verify the template is recreated, callGET /_template/watch_history
. Note that this just updates the template, so you need to wait one day for this update to take effect when a new history index is created. -
The
watcher.http.proxy.port
setting for global proxy configuration was not applied correctly.
2.3.4
editJuly 7, 2016
Bug Fixes
-
Putting a new watch with state
active=false
now stores that state correctly on a multi node cluster. -
Fixed the watch history template so nested request bodies are handled
correctly. To update an existing installation, delete the existing watch history
template by running
DELETE /_template/watch_history
and the correct template will be autocreated. To verify the template is recreated, callGET /_template/watch_history
. Note that this just updates the template, so you need to wait one day for this update to take effect when a new history index is created. - The HTML sanitizer now supports border and cellpadding attributes on table elements and the colspan and rowspan attributes on <td> and <tr> elements.
- Fixed the Watcher/Marvel examples in the documentation.
2.3.3
editMay 18, 2016
Enhancements
- Adds support for Elasticsearch 2.3.3
2.3.2
editApril 26, 2016
Bug Fixes
- All SMTP connection timeouts are now set to two minutes by default to prevent a watch from getting stuck.
-
HTTP headers from responses that contained dots led to exceptions when the HTTP response was stored in the watch history. All dots in any header names are now replaced with underscores. For example, a header called
foo.bar
becomesfoo_bar
- Hipchat action: Fall back to the default Hipchat color and format if they are not specified at the account level or within the action itself, instead of failing.
2.3.1
editApril 4, 2016
Enhancements
- Adds support for Elasticsearch 2.3.1
2.3.0
editMarch 30, 2016
New Features
- Added PagerDuty action
-
Added support for adding attachments to emails via HTTP requests and superceding and deprecating the usage of
attach_data
in order to use this feature
Enhancements
-
Support
url
in http requests as a shortcut forpath
,scheme
,port
,params
-
Support
ignore_condition
andrecord_execution
as parameters in the Execute Watch API
Bug Fixes
- The http client does not do any URL escaping by itself anymore, preventing potential wrong double escapes
2.2.1
editMarch 10, 2016
Bug Fixes
-
The
croneval
CLI tool sets the correct environment to run
2.2.0
editFebruary 2, 2016
Enhancements
- Adds support for Elasticsearch 2.2.0.
2.1.2
editFebruary 2, 2016
Enhancements
- Adds support for Elasticssearch 2.1.2
2.1.1
editDecember 17, 2015
Bug Fixes
- Fixed an issue that prevented sending of emails
2.1.0
editNovember 24, 2015
New Features
- Adds support for chaining several inputs
Enhancements
- Adds support for Elasticsearch 2.1.0.
-
Adds support for configuring a proxy in the webhook action, http input and configuring a default proxy (which is also used by the slack action), using the
watcher.http.proxy.host
andwatcher.http.proxy.port
settings.
Bug Fixes
- Fixed an issue where the scheduler may get stuck during Watcher startup. This caused no watches to ever fire.
- Fixed an issue where under specific conditions Watcher would not start if there are not finished watch executions from the previous time that watcher was running and those watch execution are unable the execute during the current start process.
2.0.1
editNovember 24, 2015
Bug fixes
- Fixed an issue where under specific conditions Watcher would not start if there are not finished watch executions from the previous time that watcher was running and those watch execution are unable the execute during the current start process.
2.0.0
editOctober 28, 2015
Breaking Changes
-
The dynamic index names support has been removed and Elasticsearch’s date math index names support should be used instead.
The only difference between Watcher’s dynamic index names support and Elasticsearch’s date math index names support is
how timezones are expressed. In Watcher this is done via node settings, in Elasticsearch the timezone is part of the
date math index names support. Only if you’re using dynamic index names with timezones in Watcher then you need to
upgrade your watches after the upgrade, otherwise your watches will work as they did before the upgrade. For example if
watcher.dynamic_indices.time_zone
setting was set to+01:00
and a watch has the following index name<logstash-{now/d}>
then after the upgrade you need to update this watch to use the following index name<logstash-{now/d{YYYY.MM.dd|+01:00}}>
.
New Features
- Added new HipChat Action
- Added new Slack Action
- Watches now have an active state. In addition, a new API was added to activate/deactivate registered watches.
- Added new array-compare, that can compare an array of values in the Watch Execution Context Model to a given value.
Enhancements
-
Watcher continuously checks if the index templates for
.watches
,.triggered_watches
and.watch_history-*
exist. Whereas before the existence of these index templates was only checked at Watcher startup time. The absence of these index templates leads to watcher data being indexed incorrectly, which then can cause Watcher to behave incorrectly. - If Watcher was stopped via the stop Watcher api and after that a master election took place then Watcher would then unexpectedly start.
-
During Watcher start up only wait for the shards of the
.watches
and.triggered_watches
indices to be available. Before Watcher also waited for the shards of the.watch_history-*
indices, which wasn’t needed. This improved time it takes for Watcher to startup. -
If
action.auto_create_index
setting has been configured then Watcher will check if the setting is too restrictive. If theaction.auto_create_index
is too restrictive then Watcher will fail during startup with a descriptive error message.
Bug Fixes
- If Watcher was installed with Shield then the Watcher index templates couldn’t be stored and could lead to Watcher behaving incorrectly. This was caused by Watcher not detecting correctly if Shield was installed.
-
Update
croneval
command line utility to properly handle whitespaces in the elasticsearch home path. - Fixed an issue where the scheduler may get stuck during Watcher startup. This caused no watches to ever fire.
- Fixed url encoding issue in http input and webhook output. The url params were url encoded twice.
1.0.1
editJuly 29, 2015
Enhancements
- Dynamic index names now support specifying a time zone to be used when computing the names of the indices. The default is UTC. Previously, the computation was fixed to always use UTC when computing the names of the indices.
Bug Fixes
- Fixed a compatibility issue with Elasticsearch 1.6.1 and 1.7.2, which were released earlier today.
1.0.0
editJune 25, 2015
Enhancements
-
Added execution time aware dynamic index names support to
index
action,search
input, andsearch
transform. - You must now explicitly specify the unit when configuring any time value. (Numeric-only values are no longer supported.)
- Cleaned up the Get Watch API response.
- Cleaned up the Stats API response.
1.0.0-rc1
editJune 19, 2015
New Features
- Added inline watch support to the Execute API
Enhancements
- Added execution context variables support.
- Email html body sanitization is now configurable.
- It is now possible to configure timeouts for http requests in HTTP input and webhook actions.
1.0.0-Beta2
editJune 10, 2015
New Features
- Acking and Throttling are now applied at the action level rather than the watch level.
- Added support for multi-doc indexing to the index action.
- Added a queued watches metric that’s accessible via the Stats API.
- Added a currently-executing watches metric that’s accessible via the Stats API.
Enhancements
- The compare condition result now includes the value of each field that was referenced in the comparison.
- The Execute API now supports a default trigger event (breaking change)
-
The
watch_record
document structure in the.watch_history-*
indices has changed significantly (breaking change) -
A new internal index was introduced -
.triggered_watches
- Added support for headers in the Webhook Action result and the HTTP Input result.
- Add plain text response body support for the HTTP Input.
Bug Fixes
-
Disallow negative time value settings for
throttle_period
- Added support for separate keystore and truststore in Webhook Action and HTTP Input.