2015年12月30日 エンジニアリング

Log In, Upgrade, Be Happy

By Erik Redding

UPDATE: This article refers to our hosted Elasticsearch offering by an older name, Found. Please note that Found is now known as Elastic Cloud.

There are a few different types of users on Found, the official Elasticsearch as a Service by Elastic. We have customers that want the latest and greatest and are first to jump on the latest releases of Elasticsearch. On the other hand, we also have customers that want to keep things where they are because they’re satisfied with the cluster’s feature set. Either way, we’re glad you’re on board but we want to make sure you’re getting the most from your cluster.

Upgrades are Easy on Found

One of the best things about using Found is the ability for customers to be able to upgrade Elasticsearch versions and cluster sizing with just a couple of clicks. Another great feature of Found is the same-day availability of the latest Elasticsearch releases. We bill by the hour, so spinning up a test cluster is affordable.

As an Ops guy, the features on Found make me smile because I can make sure my Elasticsearch clusters are running with the latest bug fixes, security patches and performance gains in Elasticsearch. At Elastic we use Found in a myriad of ways, from site search to BI and analytics. The fast upgrade paths in combination with high-availabilty configurations let me rest easy while we dog-food Found clusters to support the running of our business.

When I’m looking to upgrade a cluster to a new release, I’ll typically build out a quick checklist of tasks. Here’s my abbreviated minor-version upgrade checklist:

  • Check that I’ve got a successful snapshot for my source cluster
  • Create a test cluster with the latest Elasticsearch release
  • Restore a snapshot from my source cluster
  • Check the logs in the Found console for nasty messages
  • Run some test queries
  • Destroy the test cluster

If everything goes well, we’ll carry out the upgrade. Major release versions typically require much more testing, but again, smoke-testing a major release upgrade can be just as easy depending on your configuration. The Found Documentation on Snapshots has some more information around renaming indexes on restore and other tips.

The Latest

Why upgrade versions, you ask? There are some key features for different Elasticsearch versions that we want our Found customers to be made aware of and embrace.

1.X Clusters

In general, clusters running anything older than 1.7.4 are encouraged to move to 1.7.4. Breaking changes are outlined in our 1.7.X Reference Documentation; most customers will find that they can upgrade easily to 1.7.4.

If you’re wondering what’s new for the 1.7 releases, there have been a series of blog posts around the release announcements for 1.7.11.7.21.7.3, and 1.7.4. These posts outline high-level notable features for each release.

2.0.X and 2.1.X Clusters

If you’re running an Elasticsearch 2.0.X or 2.1.0 cluster without replicas, please upgrade to 2.0.2 or 2.1.1 today because there is a very important fix around translog corruption that could lead to data loss. The default cluster configuration on Found is to have a single replica shard, but many people adjust settings for their use case.

Kibana Releases

We manage all the details behind the Kibana releases so you don’t have to worry about keeping those up to date. If there’s a security-related patch like we saw with the release of 4.3.1, 4.2.2 and 4.1.4, we will upgrade the Kibana instance.

Finding Out What’s New

Customers interested in keeping up with the latest releases and improvements in Elasticsearch can follow the Elastic blog for the latest release announcements and more. The Elastic Discuss forums are a great way to keep up with the community and include security announcements.

In Conclusion…

We encourage users to upgrade to the latest patch release available for the major version of your cluster. There’s a lot going on with Elasticsearch and the latest releases bring some important stability and bug fixes. Leveraging Found features such as snapshot restoration and quick cluster provisioning can help you plan ahead for patch-set or major-version upgrades.