Graphite

From Wikitech
Jump to navigation Jump to search

Graphite is a real-time time series data store and graph renderer.

Front-ends

Wikimedia Foundation deploys various web applications that provide convenient ways to access the data or generate graphs.

  • grafana.wikimedia.org, a frontend for flexibly querying metrics and creating new graphs and dashboards. Unlike other front-ends, this queries the raw data and renders its (interactive) graphs client-side.
  • graphite.wikimedia.org (restricted), the default graphite-web frontend. Provides a visual interface to discover functions and metric names, with an API to create server-side rendered graphs in PNG format, as well as raw data in JSON format.

Service operation

The Graphite receiver for production is hosted on graphite1001. (Previously on tungsten.)

Always use graphite-in.eqiad.wmnet as the inbound receiver endpoint for graphite/carbon protocol traffic (i.e. port 2003). This decouples the inbound receiver service from the actual hosting machine allowing safer maintenance operations as well as easier HA/Load balancing. For statsd pushing via udp on port 8125 see the guidelines at statsd.

For Wikimedia Cloud VPS and the Beta Cluster hosted there, the receiver is labmon1001, with frontends https://graphite-labs.wikimedia.org and https://grafana-labs.wikimedia.org. Note that metric names starting with the name of a Cloud VPS project followed by a dot are reserved for server monitoring metrics (collected with Diamond). Other metrics in this namespace may be automatically deleted without notice. For example, if your project is called "foo", use foo_extrametrics.bar_baz instead of foo.bar_baz.

Metric creation

Graphite does not have a procedure for creating or registering metrics. Instead, Graphite clients can submit data for any metric by name, and the metric is created automatically in storage backend.

Everything stored in graphite has a path with components delimited by dots. In a path such as foo.bar.baz, each segment surrounded by dots is called a path component. So "foo" is a path component, as well as "bar", etc. When coming up with metric names, adhere to the following guidelines:

  • Each path component should have a clear and well-defined purpose.
  • Volatile path components should be kept as deep into the hierarchy as possible.

The automatic creation of metrics avoids overhead of configuration, but also means metric name fields should not be publicly exposed to user input. This is instead the responsibility of Graphite clients. For a list of Graphite clients in use at Wikimedia Foundation, see #Data sources.

Terminology

  • Metric (also known as Bucket). Each metric has a name and a bucket with one or more values over time.
  • Flush interval. At a configured interval, the Statsd server will aggregate all buckets and send the representative values for each property to Graphite. At Wikimedia the interval is currently one minute.
  • Aggregation. Each minute, Statsd takes each bucket and reduces its values to a single value to represent that minute. It also creates the derivative properties at this point (e.g. "median", "rate", "sum", "p95", etc.).
  • Retention and Resolution. In Graphite, each metric has multiple databases with differing retentions and resolutions. For example, data kept for 7 days has a resolution of 1 data point per minute. Data kept for 30 days has a resolution of 1 data point per 15 minutes.[1]

Data sources

Graphite is one of the primary aggregators for metrics at Wikimedia Foundation. It providers a powerful API to query, transform, and aggregate the data.

Data is usually not submitted to Graphite directly. Instead, it should go through one of the below clients. The most commonly used client is Statsd.

statsd

The Statsd server acts as an intermediary between Graphite and other applications.

Data submitted via Statsd uses the raw metric name from Statsd as-is inside Graphite. No prefix or namespace is added by default. However, Statsd does add #Extended properties to your metrics.

EventLogging

To aggregate data from EventLogging events from client-side JavaScript, we usually create a Python deamon that subscribes to relevant EventLogging topics in Kafka, that reacts by sending packets to Statsd. See Webperf for examples. See EventLogging for how to create new schemas.

statsv

statsv is an HTTP beacon endpoint (/beacon/statsv) for sending data to Graphite via statsd.

Data pipeline (simplified):

HTTP -> Varnish -> Kafka -> statsv.py -> statsd -> Graphite

It's a lightweight way of sending data from clients. This is useful when you only require one or more values to be aggregated, without needing the overhead of an EventLogging schema or storing each entry in a relational database.

You can hit this endpoint directly with an HTTP request.

Within MediaWiki, you should use the abstraction layer provided by the WikimediaEvents extension. Like so:

mw.track( 'timing.MediaWiki.foo_bar', 1234.56 ); // milliseconds
mw.track( 'counter.MediaWiki.foo_quux', 1 ); // increment

Use the "timing" and "counter" topic namespace of mw.track (documentation). Note that a prefix is not automatically added. So for metrics relating to MediaWiki core or a MediaWiki extension, be sure to prefix under "MediaWiki." (per $wgStatsdMetricPrefix ). Likewise, for using statsv from other applications (WebPageTest, EventLogging, and Zuul are examples), be sure to include a namespace for the application.

MediaWiki

Use MediaWiki's MediaWikiServices::getInstance()->getStatsdDataFactory() interface. This buffers data within the process, and sends it to Statsd at the end.

Note that properties from MediaWiki automatically get the MediaWiki. prefix added to the metric name (configurable with $wgStatsdMetricPrefix).

TCP

To record data to Graphite directly, the client must send a message over TCP to port 2003 that will contains three space separated entries:

  1. Metric name.
  2. Integer value.
  3. Unix timestamp.

Example:

$ echo "example.metric 1911 $(date +%s)" | nc -q0 graphite-in.eqiad.wmnet 2003

Extended properties

This is the missing manual about aggregation by Statsd and Graphite, at Wikimedia. This describes the primary metric types we use: counters and timers. For other metric types, see Statsd Documentation: Metric Types.

Counters

A simple counter that resets after each interval. Aggregation layers will combine values using sum().

A single push can increment the counter with any positive number. Incrementing by 1 is most common in application code, however aggregation may happen within your application or at other layers in-between the application and Graphite. As such, Statsd may see your increments as higher than 1.

Recommended properties:

  • rate: The total per second. This is initially computed by dividing the minute's sum by 60.
    This is aggregated by Graphite using avg(). It remains accurate as "average rate per second".
    • Tip: Use scale() to plot a counter for other intervals. For example, to draw the rate per minute, use metric.rate and scale(60).
  • sum: The total per variable aggregation window. This is not always per-minute.
    This is aggregated by Graphite using sum(). Viewing older data shows higher values than recent data.
    When querying recent data only (< 7 days) points show a sum per minute. When including older data, there is only a sum() over larger windows of time (eg. 15min, 1 hour, or more).
    • Tip: Only use sum when intending to plot an accurate total. For example, in conjunction with integral().
    • Tip: To plot the rate per minute, or per second, use rate instead!
  • lower: The lowest single increment command per variable aggregation window.
    This is aggregated by Graphite using min().
    Viewing older data may show lower values than recent data due to aggregation.
  • upper: The highest single increment per variable aggregation window.
    This is aggregated by Graphite using max().
    Viewing older data may show higher values than recent data due to aggregation.

Discouraged properties:

  • count: Number of StatsD commands received per variable aggregation window. If there is no aggregation in your application or elsewhere between the application and Statd, and if all increments are by 1, then this is usually equal to sum . Otherwise, this will differ. For example, after a sequence of "+1, +4, +2" the count is recorded as 3 commands, where sum would record 7.
    This is aggregated by Graphite using sum(). Viewing older data shows higher values than recent data, see sum for why.
  • mean: The average of all increments per variable aggregation window.
    This is aggregated by Graphite using avg(). Data older than 7 days is not statistically meaningful (average of averages).

Timers

Track the duration of a particular event. Recommended properties:

  • median: The middle timing value per variable aggregation window. See also Comparison of mean and median on Wikipedia.
    This is aggregated by Graphite using avg(). Data older than 7 days is not statistically meaningful (average of average of medians).
  • sample_rate: The total number of values per second. This behaves identical to Counter.rate.
    This is aggregated by Graphite using avg(). It remains accurate as "average rate per second".
    • Tip: Use scale() to plot a counter for other intervals. For example, to draw the rate per minute, use metric.rate and scale(60).
  • p75, p95, p99, etc.: See also Percentile on Wikipedia.
    This is aggregated by Graphite using avg(). Data older than 7 days is not statistically meaningful (average of average of percentile).
  • sum: The total sum of timing durations in this interval.
    Use this to compute the (globally) total amount of time spent in your metric.
    For example "250ms, 300ms, 550ms" produces 1100ms, not 3.
    This is aggregated by Graphite using sum(). Viewing older data shows higher values than recent data.
  • lower: The lowest value in an interval. This is aggregated by Graphite using min().
  • upper: The highest value in an interval. This is aggregated by Graphite using max().

Discouraged properties:

  • count: Number of StatsD commands received per variable aggregation window. This is not reliably per minute.
    This is aggregated by Graphite using sum(). Viewing older data shows higher values than recent data.
    • Tip: Use sample_rate instead if you want a rate per known interval. For example, rate per second, or rate per minute.
  • rate: Do not use Timer.rate! For the equivalent of Counter.rate, see Timer.sample_rate! The rate here is the sum of the timer for the reporting interval (in milliseconds) divided by the reporting interval (in seconds). In other words, it is the total time of that measurement, normalized to the second. It's weird and confusing. To draw a counter from a timing metric, use sample_rate instead.
  • mean: The average of all values in this interval. This is aggregated by Graphite using avg(). Data older than 7 days is not statistically meaningful (average of averages).

Functions

Here is a short list of common functions you should know about.

summarize

For graphs showing the history of a metric over the course of several weeks or months it can be helpful to summarise data points to a higher interval to help hide normal variation. For example, it's much easier to see a regression from ~ 10 to ~ 20 on a straight line than a line that continuously wiggles between 1 and 30. Even after aggregation into a median and application of moving average, data can still exhibit a wide variation over longer periods of time.

summarize() helps you plot very bold and spaced out data points onto a graph. For example, one value per hour, day, or week. To emphasise changes in the metric more prominently, the "staircase" line mode can be used in addition to this.

See the "History" panel on the Navigation Timing dashboard for an example of the summarize() function.

aliasByNode

For longer metric paths, this function can help shorten the labels in the legend. Its benefit over assigning labels manually with alias() is that it is automatically derived from the metric path (avoids the label from becoming incorrect when the metric path is changed). It also has the benefit of deriving short names for a series containing multiple metrics, without having to name each metric separately.

Architecture

Graphite architecture Oct 2018.svg

As of October 2018 the main components of the Graphite stack are:

  • carbon frontend relay
  • carbon local relay
  • global statsd aggregator
  • HTTP API via graphite-web

statsd ingestion

statsd UDP traffic is ingested through the statsd.eqiad.wmnet DNS name on port 8125, this is all statsd traffic that needs global aggregation and by far the most widespread protocol for pushing metrics at WMF. Statsd metrics sent there to the global aggregato will produce values aggregated across all hosts that sent said metrics. For example mediawiki statsd metrics are aggregated globally, i.e. the metric name doesn't contain the host name sending the metric.

For statsd metrics where global aggregation is not desired or needed the recommended approach is to run a local statsd aggregator (in our case statsite) and point the application to localhost:8125 instead. Such cases are for example Swift metrics, any application that generates metrics including the hostname is a candidate for a local statsd aggregator. The results are going to be the same as if the metrics were aggregated globally while putting less stress on the global aggregator.

Note that if an application sends a mixture of hostname-specific metrics and global metrics then it should use statsd.eqiad.wmnet.

After aggregation is done, either local or global, the resulting metrics are flushed to Graphite using the Carbon protocol over TCP.

carbon ingestion

The Graphite protocol is called Carbon and it is TCP-based and line oriented, metrics have a name, a value and a timestamp. No metric data types are possible like in statsd. The entry point for carbon traffic is graphite-in.eqiad.wmnet DNS name on port 2003, said traffic will hit the carbon frontend relay component, implemented by the carbon-c-relay software. Once the frontend has accepted the metrics, they will be mirrored to all datacenters where graphite is present (eqiad and codfw as of October 2018) and received by the carbon local relay (also implemented by carbon-c-relay) for storage on disk.

HTTP read-only API

Once stored on disk the metrics are served by graphite-web as a uwsgi application via HTTP. The web application is what powers the Grafana backend for Graphite and graphite.wikimedia.org, it will query for metrics from all graphite hosts local to its datacenter and serve the resulting data.

FAQ

How do I render a counter metric as running total?

Start the sum property which stores the total per interval (e.g. minute or hour). Then apply integral() to produce a running total. (Original discussion at T108480)

Why is the data so different when zooming out or moving a week back?

This is most likely a side-effect from a graph using .count or .sum where .rate should be used instead. See Extended properties for how to resolve this.

Operations manual

Failover

When needing to failover traffic from one graphite host to another it is important to think about such traffic in terms of read and write traffic.

Read traffic: all reads happen via HTTP and graphite is fronted by Varnish, thus a change like 454876 is sufficient to switch read traffic.

Write traffic: write traffic is trickier, as it involves changing multiple entry points. Additional complication is brought by the fact that some clients will ignore DNS records changes and thus will need to be restarted.

  • Diamond will require a puppet change like 454874 and will write to new graphite host at the next puppet run.
  • The DNS change involved is for statsd and carbon traffic, like 454872

Monitor graphite queries

The graphite web application graphite-web does a fair amount of logging, specifically inside /var/log/graphite-web/metricaccess.log and /var/log/graphite-web/rendering.log. All requested queries are logged together with how much time it spent serving those.

Deleting metrics

There's no (as of Nov 2016) formal/periodic clean up of old or unwanted/unneeded metrics. To get a metrics deleted please file a Phabricator task under #Graphite.

For people with access to graphite: to delete metrics it is sufficient to find the correct files/directories under /var/lib/carbon/whisper from graphite1004 and graphite2003 and delete them. (Exception for cassandra. metrics, graphite1003 and graphite2002)

rsync metrics

Graphite machines run an rsync server to make metrics accessible to other graphite machines in case manual sync is needed (e.g. after machine failure). To rsync in parallel, first sync only top-level directories and then the contents in parallel:

 install -d -o _graphite -g _graphite /var/lib/carbon/whisper
 su -s /bin/bash _graphite
 cd /var/lib/carbon/whisper
 rsync -vd SOURCE::carbon/whisper/ .
 /usr/bin/time parallel -j5 -i rsync -a SOURCE::carbon/whisper/{}/ {}/ -- *

Merge and sync metrics

The rsync method above provides a "point in time" sync. Another possibility albeit slower is to sync and merge metrics from a machine onto another one. Thanks to carbonate this is possible to do in the background, where metrics are rsync'd from an host and then merged onto existing ones. Using this method the metrics files are locked during merge and thus can it be used to fill "holes" even when the destination machine is actively being written to. The carbonate package is needed (available internally on stretch) together with time and GNU parallel.

 export CARBONATE_CONF=/etc/carbon/carbonate.conf
 (cd /srv/carbon/whisper ; find . -type f | carbon-path -r) | time parallel --files --jobs 10 --pipe --block 20k -- carbon-sync -s SOURCE -d /srv/carbon/whisper --source-storage-dir :carbon/whisper -l

List slow queries

Graphite's web application logs how long a given query took, you can list all queries taking over a certain amount of time (in seconds) with:

 awk -Ftook '$2 > 10 { print }' /var/log/graphite-web/rendering.log

Note that a slow query isn't necessary the cause of a slowdown, it might be a consequence

List metrics being created

 tail -F /var/log/carbon/carbon-cache@*/creates.log | grep 'matched schema'

Operations troubleshooting

carbon-cache too many creates

This alert is used to signal whenever too many files (and therefore disk space) are being created on disk. It can be benign, e.g. when a new cassandra instance gets bootstrapped there is a flood of new metrics being created. To check which files are being created:

 sudo tail -F /var/log/upstart/carbon_cache-*.log /var/log/carbon/carbon-cache@*/creates.log | grep 'creating database file'

It is also useful to tally which metrics have been created according to "top level" (i.e. the leftmost component)

 sudo grep 'aggregation schema' /var/log/carbon/carbon-cache@*/creates.log* | awk '{print $6}' | cut -d. -f1 | sort | uniq -c | sort -rn | less

Applying carbon storage_aggregation changes

The current settings are only used when creating new Whisper data sets. Existing ones will generally not be affected. To apply, say, newer xFilesFactor configuration to an existing property, use the following steps.

While there is no script to read the current would-be settings and apply them, there is a script to manually apply specific settings.

  1. Get new xFilesFactor settings for the relevant metric from puppet:/role/graphite/base.pp. For example: "0.01" or "0".
  2. Get new retention settings from puppet:/role/graphite/base.pp. For example: "1m:7d,5m:14d,15m:30d,1h:1y,1d:5y".
  3. Check current settings:
    $ whisper-info mw/js/deprecate/tipsy_live/sum.wsp
    xFilesFactor: 0.0
  4. $ sudo -su _graphite (or whichever user is the owner of the wsp file)
  5. Run whisper-resize and set xFilesFactor, then the path, and then the retention values as distinct space-separated command-line arguments:
    $ whisper-resize --xFilesFactor=0 mw/js/deprecate/tipsy_live/sum.wsp 1m:7d 5m:14d 15m:30d 1h:1y 1d:5y
  6. Remember to run it on both Eqiad and Codfw primary Graphite hosts (as of July 2018: graphite1001 and graphite2003).

Identifying heavy and/or expensive queries

Graphite might suffer from heavy CPU or memory load if queries requesting a lot of data are run. It is possible to identify those after the fact by asking uwsgi for big (as in bytes) or long (as in milliseconds) queries by awk-ing the request logs. See also bug T116767 e.g.

 # filter for >5MB responses
 | awk '$26 > 5000000 {print }'  | less 
 # filter for > 9s responses
 | awk '$29 > 9000 {print }'

Further reading

External links

References

  1. Graphite configuration, Wikimedia operations puppet