80+ ELK Stack Interview Questions and Answers (2027)

84 ELK Stack interview questions with answers on Elasticsearch, Logstash, Kibana and Beats, grouped for freshers, mid-level, senior and scenarios.

Sep 16, 2025 - 17:27
Nov 6, 2026 - 00:32
 0  218
80+ ELK Stack Interview Questions and Answers (2027)

ELK Stack interviews test whether you can explain how Elasticsearch, Logstash, Kibana and Beats fit together, then run and fix a logging cluster: shards and replicas, mappings, data streams, ILM, Logstash pipelines, Query DSL and KQL, cluster health and security. This guide answers 84 distinct questions, grouped by experience level.

This is the main ELK question bank in our DevOps interview questions hub. For deeper incident drills, read the scenario-based ELK Stack interview questions, for performance and internals go to advanced ELK Stack questions, and for a study plan see the ELK Stack interview preparation guide. Product details were checked against Elastic's official documentation in November 2026.

Key takeaways

  • Know each component's job and default port: Elasticsearch 9200 and 9300, Kibana 5601, the Beats input on Logstash usually 5044.
  • Explain shards, replicas and cluster health colours clearly. Red means a primary shard is unassigned; yellow means a replica is.
  • For logs, modern answers use data streams, index templates and ILM rather than hand-made daily indices.
  • Be able to write a Logstash pipeline, a grok pattern, a KQL query and a bool query from memory.
  • Senior candidates should cover master quorum, heap sizing, disk watermarks, mapping explosions, tiered retention and rolling upgrades.
  • Security has been on by default for new installs since 8.0. Never expose 9200 or 5601 to the internet.

Components at a glance

ComponentRoleDefault portTypical interview focus
ElasticsearchStore, index and search9200 (HTTP), 9300 (transport)Shards, mappings, queries, cluster health
LogstashCollect, parse, enrich, route5044 (Beats input, by convention), 9600 (API)Pipelines, grok, queues, performance
KibanaExplore, visualise, manage5601Discover, Lens, dashboards, alerting
Beats / Elastic AgentShip data from hostsNone (outbound)Filebeat, DaemonSets, Fleet policies

How the questions are organised

SectionQuestionsLevel
Freshers1 to 28Components, vocabulary, first commands
2 to 5 years29 to 55Data streams, ILM, pipelines, queries
Senior56 to 70Architecture, stability, security, upgrades
Scenarios71 to 84Troubleshooting walk-throughs

ELK Stack interview questions for freshers

These cover the components, core vocabulary and the first commands you would run. Keep answers short and name the default port or API where it helps.

1. What is the ELK Stack?

ELK stands for Elasticsearch, Logstash and Kibana. Elasticsearch stores and searches data, Logstash collects and transforms it, and Kibana visualises it. Elastic now calls the wider family, including Beats and Elastic Agent, the Elastic Stack.

2. What is the difference between the ELK Stack and the Elastic Stack?

ELK names the original three products. The Elastic Stack adds Beats (lightweight shippers), Elastic Agent with Fleet for central agent management, and features such as APM and security analytics. In interviews the two names are often used interchangeably.

3. What is Elasticsearch?

Elasticsearch is a distributed search and analytics engine built on Apache Lucene. It stores JSON documents, indexes them for full-text and structured search, and scales by spreading data across shards on several nodes. You talk to it over a REST API.

4. What is Logstash?

Logstash is a server-side data processing pipeline. Each pipeline has inputs (Beats, files, Kafka, syslog), filters (grok, dissect, mutate, date) and outputs (usually Elasticsearch). It runs on the JVM.

5. What is Kibana?

Kibana is the web UI for the Elastic Stack. You use it to search data in Discover, build visualisations with Lens, assemble dashboards, manage index lifecycle and security, and run tools such as Dev Tools for API calls.

6. What are Beats?

Beats are lightweight single-purpose shippers written in Go. Filebeat ships log files, Metricbeat ships system and service metrics, Packetbeat captures network data, Heartbeat checks uptime, Auditbeat collects audit data and Winlogbeat reads Windows event logs.

7. What is the difference between Filebeat and Logstash?

Filebeat is a light agent that tails files on each host and forwards events, with modest processing. Logstash is a heavier central processor with many plugins for parsing, enrichment and routing. A common design is Filebeat on hosts sending to Logstash or straight to Elasticsearch with ingest pipelines.

8. What is Elastic Agent and Fleet?

Elastic Agent is a single agent that can collect logs, metrics and security data using integrations, replacing several Beats on one host. Fleet, in Kibana, enrolls and manages agents centrally with policies, so you change collection settings without logging in to each server.

9. What default ports does the stack use?

Elasticsearch serves its REST API on 9200 and uses 9300 for node-to-node transport. Kibana listens on 5601. Logstash's Beats input is conventionally configured on 5044, and the Logstash monitoring API listens on 9600. Restrict these ports to trusted networks.

10. What is an index in Elasticsearch?

An index is a logical collection of documents with a mapping that defines field types. For logs, you usually write to time-based indices or a data stream rather than one ever-growing index.

11. What is a document?

A document is a JSON object stored in an index, with a unique _id. Each log event typically becomes one document.

12. What are shards and replicas?

An index is split into primary shards so data and work spread across nodes. Each primary can have replica shards on other nodes for high availability and extra read capacity. Replica count can change at any time; the primary count is fixed at creation unless you use the split, shrink or reindex APIs.

13. What is a node, and what node roles exist?

A node is one running Elasticsearch instance. Roles include master-eligible (cluster state), data (with tiers such as data_hot, data_warm, data_cold, data_frozen, or the generic data role), ingest (runs ingest pipelines), coordinating-only (routes requests), ml and transform. Large clusters separate roles across dedicated nodes.

14. What is a cluster?

A cluster is one or more nodes sharing the same cluster.name that together hold all the data. One elected master node manages cluster state such as index metadata and shard allocation.

15. What do the cluster health colours mean?

Green means all primary and replica shards are assigned. Yellow means all primaries are assigned but at least one replica is not, which is normal on a single-node cluster. Red means at least one primary shard is unassigned, so some data is unavailable.

16. How do you check cluster health?

Call the cluster health API or the cat APIs from Dev Tools or curl.

GET _cluster/health
GET _cat/nodes?v
GET _cat/indices?v&s=store.size:desc
GET _cat/shards?v

17. What is an inverted index?

An inverted index maps each term to the list of documents that contain it, like the index at the back of a book. Lucene builds one per text field, which is why full-text search is fast.

18. What is a mapping?

A mapping defines how fields are stored and indexed: text, keyword, date, long, ip, boolean, geo_point and so on. Elasticsearch can infer mappings dynamically, but explicit mappings avoid wrong types in production.

19. What is the difference between text and keyword fields?

A text field is analysed (tokenised and lowercased) for full-text search. A keyword field stores the exact value for filtering, sorting and aggregations, such as a hostname or status code. Dynamic mapping often creates both, with field.keyword as a sub-field.

20. What is an analyzer?

An analyzer turns text into terms: character filters clean input, a tokenizer splits it, and token filters lowercase, remove stop words or stem. The standard analyzer is the default for text fields.

21. What is the grok filter?

Grok parses unstructured text into fields using named patterns, for example %{IP:client} %{WORD:method} %{URIPATHPARAM:request}. It ships with many patterns for common log formats.

filter {
  grok {
    match => { "message" => "%{COMBINEDAPACHELOG}" }
  }
}

22. What does a basic Logstash pipeline config look like?

Three blocks: input, filter, output. This one receives events from Beats, parses the date and sends to Elasticsearch over HTTPS.

input {
  beats { port => 5044 }
}
filter {
  date { match => ["timestamp", "ISO8601"] }
}
output {
  elasticsearch {
    hosts => ["https://es01:9200"]
    data_stream => "true"
    user => "${ES_USER}"
    password => "${ES_PASS}"
  }
}

23. What is the date filter used for?

It parses a timestamp from the log line and sets @timestamp to it. Without it, events get the time Logstash received them, which skews searches when logs arrive late.

24. What is Kibana Discover?

Discover is where you search and browse raw documents for a chosen data view and time range, add field columns, filter by values and save searches for reuse in dashboards.

25. What is a data view in Kibana?

A data view (called an index pattern before Kibana 8.0) tells Kibana which indices or data streams to query, for example logs-*, and which field is the time field.

26. What query languages can you use in Kibana?

KQL (Kibana Query Language) is the default in the search bar, Lucene query syntax is still available, Query DSL is the JSON language of the search API, and ES|QL is a piped query language for exploring and aggregating data. EQL serves event sequences in security use cases.

27. Write a KQL query for 5xx errors from one service.

KQL uses field: value pairs with boolean operators and ranges.

service.name : "checkout" and http.response.status_code >= 500

28. Why use the ELK Stack for centralized logging?

Logs from many servers, containers and services land in one searchable place, so you can correlate an error across systems, keep history after pods are deleted, build dashboards and alert on patterns instead of SSHing into hosts.

ELK Stack interview questions for 2 to 5 years of experience

Mid-level questions test hands-on work: data streams, ILM, Logstash pipelines, Beats on Kubernetes, Query DSL and Kibana alerting.

29. What is a data stream?

A data stream is an append-only abstraction over a series of hidden backing indices, designed for time series such as logs and metrics. Documents must have an @timestamp field. Rollover creates a new backing index automatically, and you query the stream by its name.

30. What is Index Lifecycle Management (ILM)?

ILM automates index management through phases: hot (active writes), warm (read-mostly), cold (rarely queried), frozen (searchable snapshots) and delete. A policy defines when to roll over and when to move, shrink, force merge or delete data.

PUT _ilm/policy/logs-30d
{
  "policy": {
    "phases": {
      "hot":    { "actions": { "rollover": { "max_primary_shard_size": "50gb", "max_age": "1d" } } },
      "warm":   { "min_age": "7d",  "actions": { "shrink": { "number_of_shards": 1 }, "forcemerge": { "max_num_segments": 1 } } },
      "delete": { "min_age": "30d", "actions": { "delete": {} } }
    }
  }
}

31. What is an index template?

A composable index template applies settings, mappings and an ILM policy to any new index or data stream whose name matches its pattern. Component templates hold reusable blocks that index templates combine.

32. What is rollover?

Rollover starts a new backing index when the current one meets a condition such as age, size or document count. It keeps shards at a healthy size and makes time-based deletion cheap.

33. How do you size shards?

Elastic's guidance is to aim for shards of roughly 10 GB to 50 GB and to keep the number of shards per node reasonable, since every shard costs heap and overhead. Use rollover on size rather than one index per day for small sources, which creates many tiny shards.

34. What is an ingest pipeline?

An ingest pipeline runs processors (grok, dissect, date, rename, geoip, set and more) on an ingest node before indexing. It lets you parse data inside Elasticsearch without running Logstash.

35. When would you choose Logstash over an ingest pipeline?

Choose Logstash when you need many input types (Kafka, JDBC, syslog), complex conditional routing to several outputs, heavy enrichment, or buffering with persistent queues. Ingest pipelines suit simpler parsing when Beats or Elastic Agent ship straight to Elasticsearch.

36. What is the dissect filter and how is it different from grok?

Dissect splits a string by fixed delimiters, so it is faster and simpler than grok but cannot handle variable layouts. Use dissect for consistently structured lines and grok when you need regex flexibility.

37. How do Logstash persistent queues work?

With queue.type: persisted, Logstash writes events to disk between input and filter stages, so events survive a restart or crash and the queue absorbs bursts. The default memory queue loses in-flight events if the process dies.

38. What is the Logstash dead letter queue?

When enabled (dead_letter_queue.enable: true), events that the Elasticsearch output cannot index, for example because of a mapping conflict, are written to a DLQ on disk instead of being dropped. You can reprocess them later with the dead_letter_queue input.

39. How do you run several pipelines in one Logstash instance?

Define them in pipelines.yml with a pipeline.id and path.config each. Separate pipelines isolate workloads, and pipeline-to-pipeline communication can split routing from processing.

- pipeline.id: nginx
  path.config: "/etc/logstash/conf.d/nginx.conf"
- pipeline.id: app
  path.config: "/etc/logstash/conf.d/app.conf"
  pipeline.workers: 4

40. What is the Elastic Common Schema (ECS)?

ECS is a shared field naming convention, such as host.name, source.ip and http.response.status_code. Mapping data to ECS lets dashboards, detection rules and integrations work across sources without custom field names.

41. How does Filebeat avoid losing or duplicating log lines?

Filebeat records how far it has read each file in its registry and resends events until the output acknowledges them, giving at-least-once delivery. Duplicates are possible after failures; setting a deterministic document ID avoids duplicate documents.

42. How do you collect logs from Kubernetes?

Run Filebeat or Elastic Agent as a DaemonSet that reads container logs from the node's log directory and adds Kubernetes metadata (pod, namespace, labels). Applications should log to stdout and stderr, ideally as JSON.

43. What is the difference between a query and a filter context?

In query context, Elasticsearch scores how well documents match (relevance). In filter context, it only answers yes or no, skips scoring and can cache results. Put exact matches and time ranges in filter clauses of a bool query.

44. Write a bool query for errors in the last 15 minutes.

Use must or filter clauses inside bool.

GET logs-app-*/_search
{
  "query": {
    "bool": {
      "filter": [
        { "term":  { "log.level": "error" } },
        { "range": { "@timestamp": { "gte": "now-15m" } } }
      ]
    }
  }
}

45. What are aggregations?

Aggregations compute summaries over matching documents: bucket aggregations group data (terms, date_histogram, range), metric aggregations calculate values (avg, sum, percentiles, cardinality), and pipeline aggregations work on other aggregation results.

46. Why can't you aggregate on a text field by default?

Aggregations and sorting need doc values or fielddata. text fields have no doc values, and fielddata is disabled by default because it uses a lot of heap. Aggregate on the keyword version of the field instead.

47. What is the refresh interval?

A refresh makes recently indexed documents searchable. The default is every second (index.refresh_interval: 1s), which is why Elasticsearch is called near real time. Raising it during bulk loads improves indexing throughput.

48. What is the translog?

The transaction log records each indexing operation so data that has not yet been committed to Lucene segments on disk can be recovered after a crash. A flush commits segments and trims the translog.

49. What is the bulk API and why use it?

The _bulk API sends many index, update or delete operations in one request, which cuts network round trips. Logstash, Beats and clients all use it. Watch for per-item errors in the response, since the request can succeed while some items fail.

50. How do you build a dashboard in Kibana?

Create visualisations, usually with Lens (drag fields, choose a chart), save them, and add them to a dashboard with shared time range, filters and controls. Save searches from Discover as dashboard panels for raw log tables.

51. How do you alert on log patterns?

Use Kibana alerting rules, such as Elasticsearch query or log threshold rules, with connectors for email, Slack, PagerDuty or webhooks. Check which connectors your subscription level includes on Elastic's subscriptions page.

52. What security features are on by default in Elasticsearch 8 and later?

Since 8.0, security is enabled by default on new self-managed installs: TLS for the transport and HTTP layers is auto-configured, the elastic superuser gets a generated password, and Kibana enrols with a token. Role-based access control limits which users see which indices.

53. What are snapshots?

Snapshots are incremental backups of indices or the whole cluster to a repository such as S3, GCS, Azure Blob or a shared filesystem. Snapshot Lifecycle Management (SLM) schedules them and enforces retention.

54. What are searchable snapshots?

Searchable snapshots let you query data stored in a snapshot repository directly. The frozen tier mounts them with a small local cache, which keeps long retention cheap at the cost of slower queries.

55. What is cross-cluster search?

Cross-cluster search queries indices on remote clusters from a local one, using cluster_alias:index syntax. Teams use it to search regional clusters from one Kibana without copying data.

Advanced ELK Stack interview questions for senior engineers

Senior rounds focus on architecture, cluster stability, security, licensing awareness, cost and upgrades.

56. How do you design a production logging architecture on Elastic?

Agents (Elastic Agent or Filebeat) on hosts and as DaemonSets, an optional buffer such as Kafka for burst protection and replay, Logstash or ingest pipelines for parsing to ECS, data streams with ILM across hot, warm, cold or frozen tiers, dedicated master nodes (three for quorum), snapshots to object storage, and Kibana behind SSO with role-based access.

57. Why do you need three master-eligible nodes?

Master election needs a majority of voting master-eligible nodes. With three, the cluster survives losing one. Since 7.0, Elasticsearch manages the voting configuration automatically and the old minimum_master_nodes setting was removed, which prevents classic split-brain when configured with cluster.initial_master_nodes only for the first bootstrap.

58. How do you size JVM heap for Elasticsearch?

Recent versions size the heap automatically based on node roles and memory, which Elastic recommends. If you set it manually, keep -Xms and -Xmx equal, no more than half of RAM so the filesystem cache has room, and below the compressed object pointers threshold (around 32 GB, check with the startup log).

59. What are disk watermarks?

Elasticsearch stops allocating shards to a node above the low watermark (85% by default), moves shards away above the high watermark (90%), and marks indices with a shard on the node read-only above the flood stage (95%). The block lifts automatically once usage drops below the high watermark.

60. What is a mapping explosion and how do you prevent it?

It happens when dynamic mapping creates thousands of fields, for example from JSON logs with user-generated keys, which bloats cluster state and heap. Prevent it with explicit mappings, dynamic: false or strict on noisy objects, the flattened field type, and the index.mapping.total_fields.limit setting (1000 by default).

61. How do you change a field's mapping on existing data?

You cannot change the type of an existing field in place. Create a new index or update the template so the next backing index gets the new mapping, then reindex old data with the _reindex API if you need it, and switch an alias. Adding new fields is allowed.

62. What are aliases and why are they useful?

An alias is a pointer to one or more indices. Applications query or write through the alias, so you can reindex into a new index and switch the alias atomically with zero downtime.

63. How do you improve indexing throughput?

Use bulk requests of a sensible size, increase the refresh interval during heavy loads, set replicas to 0 during an initial load then add them back, avoid unnecessary fields and analysis, use fast local SSDs, and spread shards across enough data nodes. Measure with the indexing stats APIs.

64. How do you speed up slow searches?

Use filters instead of scoring queries for exact matches, narrow the time range, avoid leading wildcards and heavy scripts, aggregate on keyword fields, reduce shard counts that are too high, and use the Profile API or slow logs to find the expensive part.

65. How do you secure an Elastic deployment?

Keep security enabled with TLS on transport and HTTP, use SSO or LDAP with roles scoped by index pattern, document- or field-level security where licensed, API keys with least privilege for shippers, audit logging, and no public exposure of 9200 or 5601. Patch regularly against Elastic security announcements.

66. What is the licensing situation of Elasticsearch and how does OpenSearch relate?

In 2021 Elastic moved Elasticsearch and Kibana from Apache 2.0 to a dual SSPL and Elastic License. AWS forked version 7.10.2 as OpenSearch, which moved to the Linux Foundation's OpenSearch Software Foundation in 2024. In 2024 Elastic added AGPLv3 as a further license option for the source code. Check the license of the exact version and distribution you use.

67. What is ES|QL and when would you use it?

ES|QL is Elastic's piped query language, for example FROM logs-* | WHERE http.response.status_code >= 500 | STATS count = COUNT(*) BY service.name. It suits ad hoc investigation and aggregation in Discover and the _query API, and is easier to read than nested Query DSL for analysts.

68. How do you plan retention and cost across data tiers?

Decide how long each data type must be searchable at full speed, how long it must be retained for audit, and how often old data is queried. Keep recent data on hot SSD nodes, move older data to warm or cold, put long retention on frozen searchable snapshots in object storage, and delete what policy allows.

69. How do you upgrade a cluster with minimal downtime?

Read the release and breaking changes notes, run the Upgrade Assistant in Kibana to fix deprecations, take a snapshot, then do a rolling upgrade one node at a time, disabling replica shard allocation during each restart and re-enabling it after. Major versions have a required minimum path, so check Elastic's upgrade docs for your version.

70. How do you monitor the Elastic Stack itself?

Use Stack Monitoring, preferably shipping metrics to a separate monitoring cluster with Elastic Agent or Metricbeat, so you can still see what happened when the production cluster is in trouble. Watch heap usage, GC, indexing and search latency, rejected thread pool tasks, disk usage and unassigned shards.

Scenario-based ELK Stack interview questions

Interviewers want to hear the first command you would run and how you would narrow the cause. For a longer set of incident scenarios, see our dedicated scenario guide linked above.

71. The cluster turned red. What do you do first?

Find which primary shards are unassigned with GET _cat/shards?v&h=index,shard,prirep,state,unassigned.reason and ask why with GET _cluster/allocation/explain. Common causes are a lost node, disk above the watermark or allocation filters. Fix the cause (bring the node back, free disk) and let allocation retry with POST _cluster/reroute?retry_failed=true. Restore from snapshot only if the data is truly gone.

72. Indices suddenly became read-only and writes fail. Why?

Usually the flood-stage disk watermark was hit, and Elasticsearch applied index.blocks.read_only_allow_delete. Free disk by deleting old indices, adding nodes or adjusting ILM. Recent versions remove the block automatically when usage drops; otherwise clear the setting on the affected indices.

73. Logs from one service appear in Kibana 20 minutes late. How do you troubleshoot?

Compare event @timestamp with ingest time to find where the delay is. Check Filebeat or agent backpressure and output errors, Logstash pipeline stats on port 9600 for slow filters, Elasticsearch indexing rejections in the write thread pool, and whether the date filter parses the right timezone. A timezone mistake often looks like a fixed delay.

74. Logstash CPU is at 100% and events back up. What do you check?

Use the node stats API to find the slowest filter per pipeline. Expensive grok patterns that fail and backtrack are a common cause; anchor patterns, use dissect for fixed formats, and drop unneeded events early. Then tune pipeline.workers and pipeline.batch.size, or scale out Logstash instances.

75. You see _grokparsefailure tags on many events. What next?

Grok did not match those lines. Copy a few failed messages into the Grok Debugger in Kibana Dev Tools, adjust the pattern or add alternative patterns, and route unmatched events to a separate index so they are not lost while you fix it.

76. A heap usage alert fires on data nodes and GC pauses grow. What do you look at?

Too many shards per node, large aggregations on high-cardinality fields, fielddata on text fields, oversized bulk requests or a mapping explosion. Check _cat/shards, _nodes/stats for breakers, reduce shard count with ILM rollover by size, and fix the query or mapping causing the pressure.

77. Your Kibana dashboard times out for a 30-day range. How do you fix it?

Check which panels are slow using the request inspector, aggregate on keyword fields, lower the number of buckets, add filters to narrow data, and consider transforms or rollups that pre-aggregate daily summaries. Make sure older data on cold or frozen tiers is not queried by default if users only need recent trends.

78. A mapping conflict shows the same field as both long and keyword across indices. What happened and how do you fix it?

Different sources or app versions sent different value types, and dynamic mapping chose different types in different backing indices. Define the field explicitly in the index template, fix the source to send a consistent type, and reindex affected data if queries across the range matter.

79. Security asks you to keep firewall logs searchable for 90 days and retained for one year. Design it.

Use a data stream with an ILM policy: hot for 7 days with rollover on size, warm until 30 days, cold or frozen with searchable snapshots until 90 days, then keep snapshots in a repository with SLM retention to one year. Document the restore process for data older than 90 days and test it.

80. A developer accidentally deleted an index. How do you recover?

Restore the index from the latest snapshot that contains it using the restore API, ideally under a new name first, then verify and swap via alias. If no snapshot exists, re-ingest from the original source if it still has the data. Afterwards, restrict delete privileges with roles.

81. Filebeat on Kubernetes ships duplicate log lines after node restarts. Why?

Filebeat delivers at least once, so events not yet acknowledged before a restart are resent. Persist the registry on a hostPath so it survives pod restarts, and consider generating a document ID with the fingerprint processor to deduplicate at index time.

82. Ingest volume spikes during a major incident and the cluster starts rejecting writes. What protects you?

A buffer such as Kafka or Logstash persistent queues absorbs the burst, while Beats and agents back off and retry. Longer term, add hot-tier capacity, drop or sample noisy debug logs at the source, and set alerts on write thread pool rejections before they cause loss.

83. You must mask credit card numbers and emails before logs are stored. How?

Mask at the earliest point you control: in the application logger, in Logstash with the mutate gsub option or a fingerprint, or in an ingest pipeline with gsub or redact processors. Test patterns against real samples, and restrict raw data access with roles in case something slips through.

84. How would you migrate from a self-managed ELK 7.x cluster to a current version?

Check the supported upgrade path in Elastic's docs, run the Upgrade Assistant, fix deprecated settings and mappings, snapshot everything, upgrade to the last 7.x minor, then roll to 8.x and onward. Alternatively, build a new cluster and restore snapshots or reindex from remote. Plan for security defaults, data views replacing index patterns, and client library changes.

ELK compared with other logging options

Interviewers often ask how ELK compares with alternatives. Keep it factual: Grafana Loki indexes labels rather than full log text, which lowers storage cost but limits ad hoc full-text search; OpenSearch is the Linux Foundation hosted fork of Elasticsearch 7.10.2; managed services such as Elastic Cloud remove cluster operations work. For a broader comparison, see our guide to real-time logging tools for DevOps teams and observability tools for DevOps monitoring. The case for starting early is covered in why teams should adopt centralized logging early.

How to prepare

  1. Run a three-node cluster locally with Docker Compose using Elastic's official images, then stop a node and watch health change.
  2. Ship Nginx logs with Filebeat, parse them in Logstash or an ingest pipeline, and build a Lens dashboard.
  3. Create an ILM policy and data stream, then force a rollover and inspect the backing indices.
  4. Break things on purpose: fill a disk past the watermark, send a conflicting field type, and fix each problem with the APIs in this guide.

For structured labs, our ELK Stack training in Pune covers Elasticsearch, Logstash, Kibana and Beats, in the Kharadi classroom or live online.

Official references

Frequently asked questions

How many ELK Stack questions are in this guide?

There are 84 distinct numbered questions: 28 for freshers, 27 for 2 to 5 years of experience, 15 for senior engineers and 14 scenarios.

What does ELK stand for?

Elasticsearch, Logstash and Kibana. Elasticsearch stores and searches data, Logstash processes it and Kibana visualises it. With Beats and Elastic Agent, the family is called the Elastic Stack.

What ports does the ELK Stack use?

Elasticsearch uses 9200 for the REST API and 9300 for node transport, Kibana uses 5601, the Logstash Beats input is usually set to 5044 and the Logstash monitoring API uses 9600.

Is Logstash still needed if I use Elastic Agent?

Not always. Elastic Agent with ingest pipelines handles many log sources. Logstash is still useful for complex routing, many input types, heavy enrichment and disk-backed buffering.

What is the difference between ILM and data streams?

A data stream is the write target that spreads time series data across backing indices. ILM is the policy that rolls those indices over and moves them through hot, warm, cold, frozen and delete phases.

Is security enabled by default in Elasticsearch?

Yes, for new self-managed installs since version 8.0. TLS is configured automatically, the elastic user gets a generated password and Kibana enrols using a token.

Should I learn OpenSearch or Elasticsearch?

Learn the concepts once: indices, shards, mappings, queries and lifecycle policies apply to both, since OpenSearch forked from Elasticsearch 7.10.2. Then focus on the product your target employers run.

Where can I get hands-on ELK Stack training?

Our ELK Stack training in Pune covers Elasticsearch, Logstash, Kibana and Beats with labs, in the Kharadi classroom or as live online classes across India.

What's Your Reaction?

Like Like 0
Dislike Dislike 0
Love Love 0
Funny Funny 0
Angry Angry 0
Sad Sad 0
Wow Wow 0
Mridul I am a passionate technology enthusiast with a strong focus on DevOps, Cloud Computing, and Cybersecurity. Through my blogs at DevOps Training Institute, I aim to simplify complex concepts and share practical insights for learners and professionals. My goal is to empower readers with knowledge, hands-on tips, and industry best practices to stay ahead in the ever-evolving world of DevOps.