By Rebecca Tortell April 05, 2016
SignalFx is integrated with Couchbase, a distributed NoSQL document store that’s specialized for backing web applications using JSON documents.
We’ve created a collectd plugin that connects to Couchbase’s statistics API and collects statistics about each Couchbase cluster, node, and bucket. Check out all our integrations from the new Integrations directory.
When you use collectd and the collectd-couchbase plugin, SignalFx provides built-in dashboards displaying useful metrics about Couchbase clusters, including:
Operations per second |
Current items |
Open connections |
Resident items ratio |
Cache miss rate % |
Memory quota used |
Memory headroom |
Out of memory errors |
Ejections/sec |
Disk write queue fill/drain |
Fragmentation |
Disk space used |
Using the SignalFx built-in dashboards for Couchbase clusters as well as individual nodes and buckets, you can monitor the following important metrics:
Cluster status. On the Couchbase Nodes overview dashboard, you can see at a glance the status the nodes and buckets in a given cluster. Nodes in the cluster should be seeing balanced activity. Buckets in the cluster should each have adequate memory remaining.
CPU Utilization: This dashboard also includes a percentile distribution of CPU utilization per node, allowing quick identification of unusually hot nodes. This chart shows minimum, 10th percentile, median (50th percentile), 90th percentile, and maximum CPU utilization for each node in the cluster.
Zooming in to an individual node shows that node’s activity, cache performance, and compute resource usage.
Next, let’s take a look at bucket performance.
Cache Performance Per Bucket: Resident items ratio and cache miss rate are inversely related: as the ratio of items in this bucket that are resident in memory drops, the number of get requests that require a fetch from disk will increase.
Disk Write Queue: Couchbase persists in-memory items to disk. This graph shows items added to the disk write queue in yellow, and items successfully written in blue. When Couchbase is keeping up with disk writes, these metrics are equal and the graph is green. When the disk queue is filling faster than it can be drained, this graph shows yellow areas.
Total Items and Operations Per Second: The Couchbase Buckets overview shows activity for all buckets being monitored. Operations per second includes gets, hits, increments and decrements.
Available Memory: When memory is exhausted, new items can be stored only by ejecting old items. An attempt to store a new item in a bucket with insufficient memory headroom produces an out-of-memory error: either a “temp” error (an old item will be ejected, try again) or a “non-temp” error (this item cannot be stored at all). Any out-of-memory error is cause for concern.