logclean-job: keep only 9 newest logstash daily indices

Signed-off-by: Jakub Sokołowski <jakub@status.im>

ansible/group_vars/log-dash.yml

@@ -27,6 +27,9 @@ es_lb_data_center: do-ams3
 es_lb_cluster_name: 'nimbus-logs-search'
 es_lb_api_port: 9200
 
+logclean_es_host: 'localhost'
+logclean_es_port: '{{ es_lb_api_port }}'
+
 # JVM Memory settings
 es_lb_jvm_min_heap: 2g
 es_lb_jvm_max_heap: 2g

ansible/logs.yml

@@ -39,6 +39,8 @@
   roles:
     - role: elasticsearch-lb
       tags: elasticsearch-lb
+    - role: logclean-job
+      tags: logclean-job
     - role: kibana
       tags: kibana
     - role: kibana-logtrail

ansible/main.yml

@@ -11,8 +11,8 @@
   hosts:
     - nimbus-master
   roles:
-    - { role: origin-certs, tags: origin-certs }
-    - { role: nimbus-stats, tags: nimbus-stats }
+    #- { role: origin-certs, tags: origin-certs }
+    #- { role: nimbus-stats, tags: nimbus-stats }
 
 - name: Configure Nimbus cluster
   hosts:

ansible/roles/logclean-job/README.md

@@ -0,0 +1,37 @@
+# Description
+
+This is a systemd timer that runs daily and removes from ElasticSearch logs older than `N` days.
+
+This is done because Nimbus generates a metric shitton of `TRACE` logs.
+
+# Configuration
+
+The main configuration values are:
+```yaml
+logclean_es_host: '127.0.0.1'
+logclean_es_port: 9200
+logclean_index_regex: 'logstash-*'
+logclean_keep_indices: 120
+logclean_service_name: 'logclean-job'
+logclean_service_timeout: 60
+logclean_timer_frequency: 'daily'
+```
+For sake of security minimum for `logclean_keep_indices` is `60`.
+
+# Usage
+
+To check the timer status use:
+```
+ $ sudo systemctl list-timers logclean-job.timer
+NEXT                         LEFT     LAST PASSED UNIT               ACTIVATES
+Sat 2020-02-08 00:00:00 UTC  10h left n/a  n/a    logclean-job.timer logclean-job.service
+```
+You can check job logs using:
+```
+ $ sudo journalctl -o cat -a -u logclean-job.service
+...
+Starting "Job for cleaning ElasticSearch cluster periodically."...
+Checking ElasticSearch for indices to clean....
+Nothing to remove. (3/90 indices)
+Started "Job for cleaning ElasticSearch cluster periodically.".
+```

ansible/roles/logclean-job/defaults/main.yml

@@ -0,0 +1,13 @@
+---
+logclean_es_host: ~
+logclean_es_port: ~
+logclean_script_path: '/usr/local/bin/logclean'
+logclean_index_regex: 'logstash-*'
+logclean_keep_indices: 90
+logclean_service_path: '/lib/systemd/system'
+logclean_service_name: 'logclean-job'
+logclean_service_desc: 'Job for cleaning ElasticSearch cluster periodically.'
+logclean_service_user: 'nobody'
+logclean_service_after: 'docker.service'
+logclean_service_timeout: 60
+logclean_timer_frequency: 'daily'

ansible/roles/logclean-job/tasks/main.yml

@@ -0,0 +1,39 @@
+---
+- name: 'Log Clean | Verify we keep a minimum of 60 indices'
+  assert:
+    quiet: true
+    that:
+      - logclean_keep_indices > 60
+
+- name: 'Log Clean | Create logclean script'
+  template:
+    src: logclean.sh
+    dest: '{{ logclean_script_path }}'
+    mode: 0755
+
+- name: 'Log Clean | Create systemd service: {{ logclean_service_name }}'
+  template:
+    src: logclean.service.j2
+    dest: '{{ logclean_service_path }}/{{ logclean_service_name }}.service'
+    mode: 0644
+
+- name: 'Log Clean | Create systemd timer: {{ logclean_service_name }}'
+  template:
+    src: logclean.timer.j2
+    dest: '{{ logclean_service_path }}/{{ logclean_service_name }}.timer'
+    mode: 0644
+
+- name: Reload systemctl
+  command: systemctl daemon-reload
+
+- name: 'Log Clean | (Re)start fetching service: {{ logclean_service_name }}'
+  service:
+    name: '{{ logclean_service_name }}.service'
+    state: started
+    enabled: true
+
+- name: 'Log Clean | Enable the service timer: {{ logclean_service_name }}'
+  systemd:
+    name: '{{ logclean_service_name }}.timer'
+    state: started
+    enabled: yes

ansible/roles/logclean-job/templates/logclean.service.j2

@@ -0,0 +1,12 @@
+[Unit]
+Description="{{ logclean_service_desc }}"
+Documentation=https://github.com/status-im/infra-nimbus/tree/master/ansible/roles/esclean-job
+Requires=network-online.target
+After=network-online.target
+After={{ logclean_service_after }}
+
+[Service]
+User={{ logclean_service_user }}
+ExecStart={{ logclean_script_path }}
+Type=oneshot
+TimeoutStartSec={{ logclean_service_timeout }}

ansible/roles/logclean-job/templates/logclean.sh

@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+
+# ElasticSearch access
+ES_HOST="{{ logclean_es_host | mandatory }}"
+ES_PORT="{{ logclean_es_port | mandatory }}"
+
+# Keep only this number of newest indices
+INDICES_KEEP="{{ logclean_keep_indices | mandatory }}"
+ES_REGEX="{{ logclean_index_regex }}"
+
+ES_URL="http://${ES_HOST}:${ES_PORT}"
+
+echo "Checking ElasticSearch for indices to clean...."
+
+# Get list of indices
+INDICES=$(curl -s "${ES_URL}/_cat/indices/${ES_REGEX}?pretty&h=index&s=index")
+INDICES_NUM=$(echo "${INDICES}" | wc -l)
+
+# If there are less indices than days stop
+if [[ ${INDICES_NUM} -le ${INDICES_KEEP} ]]; then
+    echo "Nothing to remove. (${INDICES_NUM}/${INDICES_KEEP} indices)"
+    exit 0
+fi
+
+# Subtract how many to keep from number of existing indices
+INDICES_TO_DELETE=$(echo "${INDICES}" | tail -n$((INDICES_NUM-INDICES_KEEP)) )
+
+echo "${INDICES_TO_DELETE}"
+
+while IFS= read -r INDEX; do
+    echo "Deleting: ${INDEX}"
+    curl -s -XDELETE "${ES_URL}/${INDEX}"
+done <<< "${INDICES_TO_DELETE}"

ansible/roles/logclean-job/templates/logclean.timer.j2

@@ -0,0 +1,6 @@
+[Unit]
+After=multi-user.target
+
+[Timer]
+OnCalendar={{ logclean_timer_frequency }}
+Persistent=yes