Report k8s pct metrics from enrichment process #7677
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
exekias
added
enhancement
module
containers
added 3 commits
July 23, 2018 17:41
Instead of doing it from the `state_container`. Problem with the previous approach is that `state_container` metricset is not run in all nodes, but from a single point. Making performance metrics not available in all cases. With this new approach, the enriching process will also collect performance metrics, so they should be available everywhere where the module is run.
exekias
force-pushed
the
fix-k8s-pct-fields
branch
from
6c83b4f
to
0b2ee48
Compare
added 2 commits
July 23, 2018 17:42
ruflin
reviewed
Jul 23, 2018
| @@ -2,12 +2,31 @@ | |||
| "comment": "", | |||
| "ignore": "test github.com/elastic/beats", | |||
| "package": [ | |||
| { | |||
| "checksumSHA1": "GYschMPw/Zi5C4QnGKRfgljnx/c=", | |||
| "path": "github.com/gogo/protobuf/proto", | |||
There was a problem hiding this comment.
@exekias Can you check if some of these we also have in the global vendor directory? If yes, we should bring the versions in sync.
There was a problem hiding this comment.
It seems we have github.com/golang/protobuf under base vendor, which differs from this one
ruflin
approved these changes
Jul 23, 2018
tsg
pushed a commit
that referenced
this pull request
Jul 24, 2018
* Fix breaking change in monitoring data (#7563) The prefix for the stats metrics was metrics but renamed to `stats` by accident as the name is now auto generated. This reverts this change. Closes #7562 * Add http.request.mehod to Kibana log filset (#7607) Take `http.request.method` from ECS and apply it to the Kibana fileset. Additional logs are added to the example log files. * Fix rename log message (#7614) Instead of the from field the to field was logged. * Add tests to verify template content (#7606) We recently started to move fields.yml into the Golang binary to be used internally. To make sure the loading important and loading of all the data into the binary works as expected for Metricbeat, this adds some basic tests. Related to #7605. * Basic support of ES GC metrics for jvm9 (#7628) GC log format for JVM9 is more detailed than for JVM8. Differences and possible improvements: * To get cpu_times.* a corellation between log lines is required. * Some GC metrics are available in jvm8 are not in jvm9 (class_unload_time_sec, weak_refs_processing_time_sec, ...) * heap.used_kb is empty, but it can be calculated as young_gen.used_kb + old_gen.size_kb * GC phase times are logged in miliseconds vs seconds in jvm8 * Improve fields.yml generator of modules (#7533) From now on when a user provides a type hint in an Ingest pipeline, it's added to the generated `fields.yml` instead of guessing. Closes #7472 * Fix filebeat registry meta being nil vs empty (#7632) Filebeat introduces a meta field to registry entries in 6.3.1. The meta field is used to distuingish different log streams in docker files. For other input types the meta field must be null. Unfortunately the input loader did initialize the meta field with an empty dictionary. This leads to failing matches of old and new registry entries. Due to the match failing, old entries will not be removed, and filebeat will handle all files as new files on startup (old logs are send again). Users will observe duplicate entries in the reigstry file. One entry with "meta": null and one entry with "meta": {}. The entry with "meta": {} will be used by filebeat. The null-entry will not be used by filebeat, but is kept in the registry file, cause it has now active owner (yet). Improvements provided by this PR: * when matching states consider an empty map and a null-map to be equivalent * update input loader to create a null map for old state -> registry entries will be compatible on upgrade * Add checks in critical places replacing an empty map with a null-map * Add support to fix registry entries on load. states from corrupted 6.3.1 files will be merged into one single state on load * introduce unit tests for loading different registry formats * introduce system tests validating output and registry when upgrading filebeat from an older version Closes: #7634 * Heartbeat Job Validation + addition of libbeat/mapval (#7587) This commit seeks to establish a pattern for testing heartbeat jobs. It currently tests the HTTP and TCP jobs. It also required some minor refactors of those tasks for HTTP/TCP. To do this, it made sense to validate event maps with a sort of schema library. I couldn't find one that did exactly what I wanted here, so I wrote one called mapval. That turned out to be a large undertaking, and is now the majority of this commit. Further tests need to be written, but this commit is large enough as is. One of the nicest things about the heartbeat architecture is the dialer chain behavior. It should be the case that any validated protocol using TCP (e.g. HTTP, TCP, Redis, etc.) has the exact same tcp metadata. To help make testing these properties easy mapval lets users compose portions of a schema into a bigger one. In other words, you can say "An HTTP response should be a TCP response, with the standard monitor data added in, and also the special HTTP fields". Even having only written a handful of tests this has uncovered some inconsistencies there, where TCP jobs have a hostname, but HTTP ones do not. * Only fetch shard metrics from master node (#7635) This PR makes it so that the `elasticsearch/shard` metricset only fetches information from the Elasticsearch node if that node is the master node. * Create (X-Pack Monitoring) stats metricset for Kibana module (#7525) This PR takes the `stats` metricset of the `kibana` Metricbeat module and makes it ship documents to `.monitoring-kibana-6-mb-%{YYYY.MM.DD}` indices, while preserving the current format/mapping expected by docs in these indices. This will ensure that current consumers of the data in these indices, viz. the X-Pack Monitoring UI and the Telemetry shipping module in Kibana, will continue to work as-is. * Add kubernetes specs for auditbeat file integrity monitoring (#7642) * Release the rename processor as GA * Fix log message for Kibana beta state (#7631) From copy paste Kafka was in the log message instead of Kibana. * Clean up experimental and beta messages (#7659) Sometimes the old logging mechanism was used. If all use the new one it is easier to find all the entries. In addition some messages were inconsistent. * Release raid and socket metricset from system module as GA (#7658) * Release raid and socket metricset from system module as GA * remove raid metricset title * Update geoip config docs (#7640) * Document breaking change in monitoring shcema Situation: * Edit breaking changes statement about monitoring schema changes (#7666) * Marking Elasticsearch module and its metricsets as beta (#7662) This PR marks the `elasticsearch` module and all its 8 existing metricsets all as `beta`. Previously only 2 metricsets were marked as `beta` with the remaining 6 marked as `experimental`. * Increase kafka version in tests to 1.1.1 (#7655) * Add missing mongodb status fields (#7613) Add `locks`, `global_locks`, `oplatencies` and `process` fields to `status` metricset of MongoDB module. * Remove outdated vendor information. (#7676) * Fix Filebeat tests with new region_iso_code field (#7678) In elastic/elasticsearch#31669 the field `region_iso_code` was added to the geoip processor. Because of this test broke with the most recent release of Elasticsearch as the events contain an undocumented field. * Fix duplicated module headers (#7650) * Fix duplicated module headers Closes #7643 * fix metricset titles for munin and kvm * fix imssing kubernetes apiserver metricset doc * remove headers from modules / metricset generator and clean up traefik title * Release munin and traefik module as beta. (#7660) * Release munin and treafik module as beta. * fixes to munin module * Report k8s pct metrics from enrichment process (#7677) Instead of doing it from the `state_container`. Problem with the previous approach is that `state_container` metricset is not run in all nodes, but from a single point. Making performance metrics not available in all cases. With this new approach, the enriching process will also collect performance metrics, so they should be available everywhere where the module is run. * Fix misspell in Beats repo (#7679) Running `make misspell`. * Update sarama (kafka client) to 1.17 (#7665) - Update Sarama to 1.17. The Sarama testsuite tests kafka versions between 0.11 and 1.1.0. - Update compatible versions in output docs - Add compression_level setting for gzip compression * Update github.com/OneOfOne/xxhash to fix mips * Update boltdb to use github.com/coreos/bbolt fork Closes #6052 * Generate fields.yml using Mage (#7670) Make will now delegate to mage for generating fields.yml. Make will check if the mage command exists and go install it if not. The FIELDS_FILE_PATH make variable is not longer used because the path(s) are specified in magefile.go. This allows fields.yml to be generated on Windows using Mage. The CI scripts for Windows have been updated so that fields.yml is generated for all Beats during testing. This also adds a make.bat in each directory where building occurs to give Windows users a starting point. Some fixes were made to the generators because: - rsync was excluding important source files contained in a directory named "build" - the generated project needed to be `git init` before running certain magefile targets that detect project's root dir and import path. * Update go-ucfg to 0.6.1 (#7599) Update fixes config unpacking if users overwrite settings from CLI, with missing values. When using `-E key=` (e.g. in scripts defining potential empty defaults via env variables like `-E key=${MYVALUE}`), an untyped `nil`-values was inserted into the config. This untyped value will make Unpack fail for most typed settings. * Docs: Add deprecation check for dashboard loading. (#7675) For APM Server the recommended way of loading dashboards and Kibana index pattern will be through the Kibana UI from 6.4 on. Since the docs are based on the libbeat docs we need to add a deprecation flag for dashboard and index pattern related documentation. relates to elastic/apm-server#1142 * Update expected filebeat module files for geoip change
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Instead of doing it from the
state_container. Problem with theprevious approach is that
state_containermetricset is not running in allnodes, but from a single point. Making performance metrics not available
in all cases (in the DaemonSet).
With this new approach, the enriching process will also collect
performance metrics, so they should be available everywhere where the
module is run.