User guide

  • ! Date/Times are displayed in Kibana in UTC

    • Headline dates in Nagios alerts are not in UTC, but if you drill down into the alert you’ll see a UTC date/time

  • On-prem Livelk:

    • This query gets you all documents with a field called type that has a value of logstash_server_log: type:logstash_server_log
    • You are only entering the fields you want to search on, not the index

      • The index is selected on the left
      • defaults to [logstash-]YYYY.MM.DD, which means it will search all logstash indices that match that [attern - to get the last two days use the time period top right
      • You can select a different index - use the dropdown top left
      • This is actually going to search everything on the selected index
    • Default search term is wildcard (*) which will return all data on the selected index (or range of indices)

    • You can change what data is shown in your results -

      • the default is two columns, one for timestamp and one for message
      • to change this, on the left click Add for different fields
      • By default as soon as you add a new column, you lose the “message” column - but you can add it back in again in the same way on the left
      • You can get the same effect by expanding a result and clicking the little grid next to the field
    • You can filter for particular values

      • Note that with a starting point of type:logstash_server_log, you are already filtering for values of type = logstash_server_log
      • For instance if you ran the type:logstash_server_log query, but then wanted to only see documents with beat.hostname =, you can expand an individual result that has that value, and click the little plus sign in a magnifying glass
      • If you want to filter for a value not currently visible to you, you can do the same but then click the resulting green losenge (top left), and edit the value
    • You can just select documents which have a particular data field.

      • This query will return all documents which have a kafka_consumer_group field: _exists_:kafka.consumer_group
    • You can change the time range of your results - click on the time value top right

      • !Watch out! If, for instance, you go from last 15 minutes to last hour, the x axis will change from “@timestamp per 30 seconds” to “@timestamp per minute” which means your bars will go twice as high up he y axis and it will look like you have twice as much data!
      • > For custom ranges, click Relative over on the left
    • Viewing / tweaking full queries:

      • The url allows you to see what your current full query is
    • For aggregations (for instance, for all type:logstash_server_log results, tell me how many of each are from each host - see thingelk below for an example of an aggregation):

      • To start a new aggregation after you already have an existing one:

      • For a simple aggregation

        • Click Visualisation at the top of the screen
        • Select “From a new search”
        • Select a visualisation type - the obvious one being Vertical bar chart (at the bottom)
        • Select your index pattern (eg [logstash-]YYYY.MM.DD) - this is just the set of indices that contain the data you’re looking for.
        • Select what type of aggregation you want in the Y-Axis section - the most common being count, because it allows you to graph the number of things that fulfil your criteria

          • Count will be selected by default anyway - you only see it if you expand the Y-Axis section
          • Other options are things like Average, Sum, Min or Max
        • Select X-Axis

          • This will give you a separate bar along the x-axis for each group of results
          • Alternatively, select Split Bars and this will give you bars that are split into sections, one section for each group of results
        • Select Aggregation: Terms

        • Select field: beat.hostname (no need for scrolling, just type quickly and press Enter even if you don’t see the thing you just typed in)

        • The default Order By of metric:count is normally the one you want

        • !! Watch out: The default Size is 5 - what this actually means is that you’ll only the 5 most frequent results. If you want more than that, increase this number. !!

        • Type your query in the main entry field at the top (contains * by default) - eg type:mobile_info_log

        • Click the green play button (top left)

        • Remember you won’t see a time axis because there is no time dimension after you have aggregated

          • But you can add a time dimension by changing the aggregation type from Terms to Date histogram or Date Range and then adjusting the Interval - and it will group results by time range (eg a separate group of results for every minute).
          • I think if you want time you can also start again with an Area Chart and have two dimensions - one for hostname and one for timestamp (not sure how though
        • If you want to group by 2 things:

          • Eg if you want to group results by gcp_node = true and gcp_node = false, then for each of those groups, break them down by hostname - so that you can see which are the top hosts for both gcp_node = true and gcp_node = false.
          • Do the same as above, but instead of selecting field: beat.hostname, select field:gcp_node - and instead of selecting X-Axis, select Split Bars
          • This should give you a giant block which looks like it’s all made of gcp_node = true, but if you look carefully you’ll see a thin strip representing gcp_node=true.
          • Now you can click add sub-bucket and give the following info:

      • For a two-dimensional aggregation,

        • For visualisation type click Area Chart
        • This example will work: for search source, click “from a new search” with index pattern [logstash-]YYYY.MM.DD
        • Bucket type X-Axis, Aggregation = Date Histogram
        • Add sub-bucket: Split Area, Aggregation = Terms
        • Field = kafka.consumer_group

          • Warning! You have to type super fast!!
    • Extra ways of building queries:

      • We used logstash_router:logstash*ab5* in Kibana to find stuff coming from AB5
      • We used NOT from_kafka:true in Kibana to rule out stuff that was coming from kafka
    • Using AND in Kibana queries:

      • AND needs to be in caps
      • You need brackets around the whole thing if it has an AND in it
  • Thingelk

    • This query gets you all documents with a field called type that has a value of check_auditd: type:check_auditd
    • This query gets you the same as above, but this time we are aggregating across the host field, which means that we will get one result for each unique host, with a count: type:check_auditd | agg:terms field:host

      • (Scroll down past the empty graph to see results)
      • So, for instance, if there were 12 results to the original query, with 6 for each of two hosts, we’ll just get two results - one for each host - with a count of 6 for each
      • In this case, the graph will be empty

        • This is because the graph is always showing you events over time, and now that we’ve aggregated over hosts we no longer have a time element available
    • You can filter results by time and individual field values, and change what data is displayed in your results in a very similar way to livelk (see above)

Jmxtrans is a translator

  • ...between kafka and graphite

  • Kafka has its own internal mettrics which go via JVM in a standridesed JVM eay

  • JMXTRans taps into that Kafka data and converts it into a format that Graphte can understand, then Grafana reads from Graphite

  • So when you click on title in Grafana

  • You then click on Metrics and see a list of things, they add up to the Graphite path

  • If you see $environment that tells you there is a variable - then ify ou look top left you’ll see a dropdown with different values for that variable (eg acme-logging-prod)

Use json to access Kibana API

  • Instead of using the Kibana UI, you can query the Kibana API directly

  • POST to (you’ll have to fill in the day)

    • So for instance if you want to download the data that’s being returned for a particular Kibana search
    • say your search is Payments_EPMandate AND "would have"
    • Post this json (untested):

      • {
         "query": {
           "filtered": {
             "query": {
               "query_string": {
                 "analyze_wildcard": true,
                 "query": "Payments_EPMandate AND \"would have\""
    • You could also download the results of a visualisation (in the UI) as a csv