Today, we are happy to announce the release of Elasticsearch Marvel 1.1.0. This is the first feature release since the introduction of Marvel, slightly more than a month ago. We've gotten great feedback from our users thus far, and Marvel 1.1.0 includes a couple of new features based on their requests. Here are the highlights:
Elasticsearch 1.0 Support in Sense
Sense's knowledge base has been extended to incorporate new features introduced in Elasticsearch 1.0. Most notably, autocomplete suggestions are now available for both the Aggregations and Snapshot and Restore APIs. All API breaking changes have also been incorporated. These updates will make it easier to both explore these new APIs via Sense and discover how your API calls need to change when moving to 1.0.
Visual Improvements to the Nodes & Indices Overview
Changes to the Nodes & Indices section provide new visual cues to help you understand the state of your cluster. The current master node is indicated with a little star. If a node drops off the cluster and stops sending data, the corresponding row will be greyed out to indicate that the data is stale and some action is needed. The same thing happens if an index was deleted, indicating the index no longer exists.
One of the goals when building Marvel was to offer in-depth monitoring and insights into the heart of Elasticsearch. As such, Marvel's agent shipped statistics on the cluster and index levels, even all the way to shard level information - be it a primary shard or a replica. Since the release, we have learned that shard level statistics generate considerable amounts of data and some users find they do not provide sufficient value to be worth the extra resources.
Shard level statistics are useful to pinpoint bad performance to the level of a specific shard misbehaving. However, in practically all cases the unintended behavior is caused by load on the Node hosting the shard. As Marvel already makes Node level information readily available, we have decided to disable shipping of shard level statistics by default. You can still enable them, in run-time, if needed. We expect this change to result in big resource savings, especially in deployments with many indices.
To help even more, we have also increased the default sampling rate from 5 seconds to 10 seconds.
Dynamically Updatable Settings
Marvel now allows you to use the Cluster Update Settings API to change some of its settings. You can now temporarily disable data shipping and change the target to which the agent sends data without restarting the cluster. This will be very helpful during maintenance and upgrades.
For a complete change list, see the documentation.
As mentioned at the start of this post, we welcome all feedback so we can continue improving Elasticsearch Marvel. Please send questions, praise or pains to the Elasticsearch User Mailing List.