This is your weekly summary of Compose news for those changes and updates which can make your life easier. In this October 8th 2018 edition, we look at how scaling databases, portals and restarting work together and we also take a browse through the past week's Compose Articles.
Today's Noteworthy is a handy hint about how you can restart your databases if you feel you need to do so. By design, we don't actually let you press a button marked restart on your deployments. The problem with restart buttons is that they can get pressed as a solution to a problem which may not actually need a restart.
Running low on resources for your database is one example where restarting might be reached for. The database would get restarted and everything would appear to be much better for a brief interval and then all the applications and users would get back online and you'd be back at square one.
The real problem is those low resources and that's where you'd want to scale up, ahead of the Compose autoscaling algorithms. The bonus, when you scale up all Compose databases (excluding PostgreSQL) is that the scaling up also involves a rolling restart of the database members of the deployment. As the restart rolls round, connections should reconnect to the new leader automatically.
That means, you've not only done a restart, you have added extra resources to your service which may also help alleviate your performance issues.
This also works for connections, but it should be noted that applications eating connections or leaving them unclosed is an issue you will really need to fix in your applications code. That could be achieved by using connection pooling either externally, like pgBouncer or in the drivers your applications use. We don't auto-scale portals so if you are finding performance is low but you are within the documented connection limits, then it may be a good time to scale up your portals - see MongoDB, PostgreSQL, Elasticsearch, Redis, RethinkDB, RabbitMQ, etcd 3, Scylla, Compose for MySQL and JanusGraph.
So, to sum up:
- Make sure it's not the portals that need scaling up first.
- If there's a problem making you want to restart your database, there's a high likelihood it's because of resource starvation.
- Scale up to relieve the resource starvation.
- There'll be a rolling restart that accompanies that scale up.
If you are still having trouble, remember that Compose Support is a click away in the Compose web console.
Last week, we looked at the end of horizontal scaling for two very specific Compose offerings, the cleanest way to get the PostgreSQL command line tools installed and, as always, the latest news around the database business.
- Enterprise Horizontal Scaling Changes have come to Compose Enterprise with the end of support for horizontally scaled MongoDB and ScyllaDB.
- Want the quickest way to get the
psqlcommand onto your machine without installing a copy of all of PostgreSQL? Read the latest PostgreSQL Tips: Installing the PostgreSQL Client and find out how to do it on macOS, Windows and all sorts of Linux systems.
That's it for this week's Noteworthy at Compose. Onwards to next week!