fix: set sbd.service timeout based on SBD_START_DELAY

Timeout for starting the sbd.service needs to be longer than
SBD_START_DELAY, otherwise the start of the sbd.service times out and
prevents a cluster from starting

.gitlab-ci.yml

@@ -60,6 +60,7 @@ variables:
   - sbd_check_devices_count
   - sbd_defaults_disabled
   - sbd_defaults
+  - sbd_delay_start
   - sbd_needs_atb_while_atb_disabled
   - sbd_needs_atb_while_atb_enabled
 

README.md

@@ -449,11 +449,22 @@ You may take a look at [an example](#configuring-cluster-to-use-sbd).
 
 list, default: `[]`
 
-List of name-value dictionaries specifying SBD options. Supported options are:
-`delay-start` (defaults to `false`), `startmode` (defaults to `always`),
-`timeout-action` (defaults to `flush,reboot`) and `watchdog-timeout` (defaults
-to `5`). See `sbd(8)` man page, section 'Configuration via environment' for
-their description.
+List of name-value dictionaries specifying SBD options. See `sbd(8)` man page,
+section 'Configuration via environment' for their description. Supported
+options are:
+
+* `delay-start`
+  * `false` or `integer`, defaults to `false`
+  * documented as SBD\_DELAY\_START
+* `startmode`
+  * `string`, defaults to `always`
+  * documented as SBD\_STARTMODE
+* `timeout-action`
+  * `string`, defaults to `flush,reboot`
+  * documented as SBD\_TIMEOUT\_ACTION
+* `watchdog-timeout`
+  * `integer`, defaults to `5`
+  * documented as SBD\_WATCHDOG\_TIMEOUT
 
 You may take a look at [an example](#configuring-cluster-to-use-sbd).
 

tasks/shell_pcs/sbd.yml

@@ -99,6 +99,43 @@
         node_devices: "{{ ha_cluster.sbd_devices | d([]) }}"
       register: __ha_cluster_distribute_sbd_config
 
+    - name: Configure systemd timeout for SBD
+      vars:
+        __sbd_delay_start: "{{ ha_cluster_sbd_options
+          | selectattr('name', 'match', '^delay-start$')
+          | map(attribute='value') | list }}"
+      when:
+        - __sbd_delay_start | length > 0
+        - __sbd_delay_start | first | int > 0
+      block:
+        - name: Ensure /etc/systemd/system/sbd.service.d directory exists
+          file:
+            path: /etc/systemd/system/sbd.service.d
+            state: directory
+            owner: root
+            group: root
+            mode: 0755
+
+        - name: Override start timeout for SBD
+          template:
+            src: templates/override-timeout.conf
+            dest: /etc/systemd/system/sbd.service.d/override-timeout.conf
+            owner: root
+            group: root
+            mode: 0644
+          vars:
+            # Make sure the timeout is at least the default 90 seconds.
+            # The intent is to make the timeout longer if needed, not shorter.
+            # yamllint disable rule:line-length
+            timeout_value: "{{ [90,
+              (__sbd_delay_start | first | float * 1.2) | round(0, 'ceil') | int] |
+              max }}"
+            # yamllint enable rule:line-length
+
+        - name: Reload systemd service files
+          systemd:
+            daemon_reload: true
+
 - name: Get services status - detect pacemaker
   service_facts:
 

templates/override-timeout.conf

@@ -0,0 +1,4 @@
+{{ ansible_managed | comment }}
+{{ "system_role:ha_cluster" | comment(prefix="", postfix="") }}
+[Service]
+TimeoutStartSec={{ timeout_value }}

tests/tests_sbd_all_options.yml

@@ -8,7 +8,7 @@
     ha_cluster_sbd_enabled: true
     ha_cluster_sbd_options:
       - name: delay-start
-        value: 2s
+        value: 2
       - name: startmode
         value: clean
       - name: timeout-action
@@ -132,7 +132,7 @@
         - name: Check SBD config
           assert:
             that:
-              - "'SBD_DELAY_START=\"2s\"' in __test_sbd_config_lines"
+              - "'SBD_DELAY_START=\"2\"' in __test_sbd_config_lines"
               - "'SBD_DEVICE=\"{{ __test_sbd_mount.stdout }}\"'
                 in __test_sbd_config_lines"
               - "'SBD_STARTMODE=\"clean\"' in __test_sbd_config_lines"

tests/tests_sbd_delay_start.yml

@@ -0,0 +1,99 @@
+# SPDX-License-Identifier: MIT
+---
+- name: SBD with long delay-start
+  hosts: all
+  vars_files: vars/main.yml
+  vars:
+    ha_cluster_cluster_name: test-cluster
+    ha_cluster_sbd_enabled: true
+    ha_cluster_sbd_options:
+      - name: delay-start
+        value: 101
+
+  tasks:
+    - name: Run test
+      tags: tests::verify
+      block:
+        - name: Set up test environment
+          include_role:
+            name: linux-system-roles.ha_cluster
+            tasks_from: test_setup.yml
+
+        - name: Set up test environment for SBD
+          include_role:
+            name: linux-system-roles.ha_cluster
+            tasks_from: test_setup_sbd.yml
+
+        - name: Ensure SBD config file is not present
+          file:
+            path: /etc/sysconfig/sbd
+            state: absent
+
+        - name: Ensure systemd overrides for SBD are not present
+          file:
+            path: /etc/systemd/system/sbd.service.d/override-timeout.conf
+            state: absent
+
+        - name: Run HA Cluster role
+          include_role:
+            name: linux-system-roles.ha_cluster
+            public: true
+
+        - name: Slurp SBD config file
+          slurp:
+            src: /etc/sysconfig/sbd
+          register: __test_sbd_config
+
+        - name: Decode SBD config
+          set_fact:
+            __test_sbd_config_lines: "{{
+                (__test_sbd_config.content | b64decode).splitlines() }}"
+
+        - name: Print SBD config lines
+          debug:
+            var: __test_sbd_config_lines
+
+        - name: Check SBD config
+          assert:
+            that:
+              - "'SBD_DELAY_START=\"101\"' in __test_sbd_config_lines"
+              - >
+                __test_sbd_config_lines[-1]
+                == 'SBD_OPTS="-n {{ __ha_cluster_node_name }}"'
+
+        - name: Check header for ansible_managed, fingerprint
+          include_tasks: tasks/check_header.yml
+          vars:
+            __file_content: "{{ __test_sbd_config }}"
+            __fingerprint: "system_role:ha_cluster"
+
+        - name: Slurp SBD overrides
+          slurp:
+            src: /etc/systemd/system/sbd.service.d/override-timeout.conf
+          register: __test_sbd_overrides_config
+
+        - name: Decode SBD overrides
+          set_fact:
+            __test_sbd_overrides_config_lines: "{{
+                (__test_sbd_overrides_config.content | b64decode).splitlines()
+                }}"
+
+        - name: Check SBD overrides
+          assert:
+            that:
+              - "'TimeoutStartSec=122' in __test_sbd_overrides_config_lines"
+
+        - name: Check header for ansible_managed, fingerprint
+          include_tasks: tasks/check_header.yml
+          vars:
+            __file_content: "{{ __test_sbd_overrides_config }}"
+            __fingerprint: "system_role:ha_cluster"
+
+        - name: Check firewall and selinux state
+          include_tasks: tasks/check_firewall_selinux.yml
+
+      always:
+        - name: Clean up test environment for SBD
+          include_role:
+            name: linux-system-roles.ha_cluster
+            tasks_from: test_cleanup_sbd.yml