Unify check attempt data type to uint32 already used somewhere

cmd/icingadb-migrate/convert.go

@@ -724,8 +724,8 @@ type stateRow = struct {
 	StateTimeUsec       uint32
 	State               uint8
 	StateType           uint8
-	CurrentCheckAttempt uint16
-	MaxCheckAttempts    uint16
+	CurrentCheckAttempt uint32
+	MaxCheckAttempts    uint32
 	LastState           uint8
 	LastHardState       uint8
 	Output              sql.NullString
@@ -798,10 +798,10 @@ func convertStateRows(
 			HardState:         row.LastHardState,
 			PreviousSoftState: row.LastState,
 			PreviousHardState: previousHardState,
-			CheckAttempt:      uint8(row.CurrentCheckAttempt),
+			CheckAttempt:      row.CurrentCheckAttempt,
 			Output:            icingadbTypes.String{NullString: row.Output},
 			LongOutput:        icingadbTypes.String{NullString: row.LongOutput},
-			MaxCheckAttempts:  uint32(row.MaxCheckAttempts),
+			MaxCheckAttempts:  row.MaxCheckAttempts,
 			CheckSource:       icingadbTypes.String{NullString: row.CheckSource},
 		})
 

doc/04-Upgrading.md

@@ -9,12 +9,40 @@ Please apply the `1.1.2.sql` upgrade script to your database. For package instal
 `/usr/share/icingadb/schema/mysql/upgrades/` or `/usr/share/icingadb/schema/pgsql/upgrades/`, depending on your
 database vendor.
 
+As the daemon checks the schema version, the recommended way to perform the upgrade is to stop the daemon, apply the
+schema upgrade and then start the new daemon version. If you want to minimize downtime as much as possible, it is safe
+to apply this schema upgrade while the Icinga DB v1.1.1 daemon is still running and then restart the daemon with the
+new version. Please keep in mind that depending on the distribution, your package manager may automatically attempt to
+restart the daemon when upgrading the package.
+
 !!! warning
 
     With MySQL and MariaDB, a locking issue can occur if the schema upgrade is applied while the history view is
     accessed in Icinga DB Web. This can result in the upgrade being delayed unnecessarily and blocking other queries.
     Please see [unblock history tables](#unblock-history-tables) for how to detect and resolve this situation.
 
+### Upgrading the state_history Table
+
+This release includes fixes for hosts and services reaching check attempt 256. However, on existing installations,
+the schema upgrade required to fix the history tables isn't automatically applied by `1.1.2.sql` as a rewrite of the
+whole `state_history` table is required. This can take a lot of time depending on the history size and the performance
+of the database. During this time that table will be locked exclusively and can't be accessed otherwise. This means that
+the existing history can't be viewed in Icinga Web and new history entries will be buffered in Redis.
+
+There is a separate upgrade script `optional/1.1.2-history.sql` to perform the rewrite of the `state_history` table.
+This allows you to postpone part of the upgrade to a longer maintenance window in the future, or skip it entirely
+if you deem this safe for your installation.
+
+!!! warning
+
+    Until `optional/1.1.2-history.sql` is applied, you'll have to lower `max_check_attempts` to 255 or less, otherwise
+    Icinga DB will crash with a database error if hosts/services reach check attempt 256. If you need to lower
+    `max_check_attempts` but want to keep the same timespan from an outage to a hard state, you can raise
+    `retry_interval` instead so that `max_check_attempts * retry_interval` stays the same.
+
+If you apply it, be sure that `1.1.2.sql` was already applied before. Do not interrupt it! At best use tmux/screen not
+to lose your SSH session.
+
 ### Unblock History Tables
 
 !!! info

pkg/icingadb/db.go

@@ -93,8 +93,8 @@ func NewDb(db *sqlx.DB, logger *logging.Logger, options *Options) *DB {
 }
 
 const (
-	expectedMysqlSchemaVersion    = 4
-	expectedPostgresSchemaVersion = 2
+	expectedMysqlSchemaVersion    = 5
+	expectedPostgresSchemaVersion = 3
 )
 
 // CheckSchema asserts the database schema of the expected version being present.

pkg/icingadb/v1/checkable.go

@@ -30,7 +30,7 @@ type Checkable struct {
 	IconImageAlt          string       `json:"icon_image_alt"`
 	IconImageId           types.Binary `json:"icon_image_id"`
 	IsVolatile            types.Bool   `json:"is_volatile"`
-	MaxCheckAttempts      float64      `json:"max_check_attempts"`
+	MaxCheckAttempts      uint32       `json:"max_check_attempts"`
 	Notes                 string       `json:"notes"`
 	NotesUrlId            types.Binary `json:"notes_url_id"`
 	NotificationsEnabled  types.Bool   `json:"notifications_enabled"`

pkg/icingadb/v1/history/state.go

@@ -14,7 +14,7 @@ type StateHistory struct {
 	HardState          uint8           `json:"hard_state"`
 	PreviousSoftState  uint8           `json:"previous_soft_state"`
 	PreviousHardState  uint8           `json:"previous_hard_state"`
-	CheckAttempt       uint8           `json:"check_attempt"`
+	CheckAttempt       uint32          `json:"check_attempt"`
 	Output             types.String    `json:"output"`
 	LongOutput         types.String    `json:"long_output"`
 	MaxCheckAttempts   uint32          `json:"max_check_attempts"`

pkg/icingadb/v1/state.go

@@ -9,7 +9,7 @@ type State struct {
 	EnvironmentMeta           `json:",inline"`
 	AcknowledgementCommentId  types.Binary               `json:"acknowledgement_comment_id"`
 	LastCommentId             types.Binary               `json:"last_comment_id"`
-	CheckAttempt              uint8                      `json:"check_attempt"`
+	CheckAttempt              uint32                     `json:"check_attempt"`
 	CheckCommandline          types.String               `json:"check_commandline"`
 	CheckSource               types.String               `json:"check_source"`
 	SchedulingSource          types.String               `json:"scheduling_source"`

schema/mysql/schema.sql

@@ -292,7 +292,7 @@ CREATE TABLE host_state (
   hard_state tinyint unsigned NOT NULL,
   previous_soft_state tinyint unsigned NOT NULL,
   previous_hard_state tinyint unsigned NOT NULL,
-  check_attempt tinyint unsigned NOT NULL,
+  check_attempt int unsigned NOT NULL,
   severity smallint unsigned NOT NULL,
 
   output longtext DEFAULT NULL,
@@ -460,7 +460,7 @@ CREATE TABLE service_state (
   hard_state tinyint unsigned NOT NULL,
   previous_soft_state tinyint unsigned NOT NULL,
   previous_hard_state tinyint unsigned NOT NULL,
-  check_attempt tinyint unsigned NOT NULL,
+  check_attempt int unsigned NOT NULL,
   severity smallint unsigned NOT NULL,
 
   output longtext DEFAULT NULL,
@@ -1147,7 +1147,7 @@ CREATE TABLE state_history (
   hard_state tinyint unsigned NOT NULL,
   previous_soft_state tinyint unsigned NOT NULL,
   previous_hard_state tinyint unsigned NOT NULL,
-  check_attempt tinyint unsigned NOT NULL,
+  check_attempt int unsigned NOT NULL, -- may be a tinyint unsigned, see https://icinga.com/docs/icinga-db/latest/doc/04-Upgrading/#upgrading-to-icinga-db-v112
   output longtext DEFAULT NULL,
   long_output longtext DEFAULT NULL,
   max_check_attempts int unsigned NOT NULL,
@@ -1343,4 +1343,4 @@ CREATE TABLE icingadb_schema (
 ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin ROW_FORMAT=DYNAMIC;
 
 INSERT INTO icingadb_schema (version, timestamp)
-  VALUES (4, UNIX_TIMESTAMP() * 1000);
+  VALUES (5, UNIX_TIMESTAMP() * 1000);

schema/mysql/upgrades/1.1.2.sql

@@ -2,3 +2,12 @@ UPDATE icingadb_schema SET timestamp = UNIX_TIMESTAMP(timestamp / 1000) * 1000 W
 
 ALTER TABLE history ADD INDEX idx_history_event_time_event_type (event_time, event_type) COMMENT 'History filtered/ordered by event_time/event_type';
 ALTER TABLE history DROP INDEX idx_history_event_time;
+
+ALTER TABLE host_state MODIFY COLUMN check_attempt int unsigned NOT NULL;
+
+ALTER TABLE service_state MODIFY COLUMN check_attempt int unsigned NOT NULL;
+
+ALTER TABLE state_history MODIFY COLUMN check_attempt tinyint unsigned NOT NULL COMMENT 'optional schema upgrade not applied yet, see https://icinga.com/docs/icinga-db/latest/doc/04-Upgrading/#upgrading-to-icinga-db-v112';
+
+INSERT INTO icingadb_schema (version, timestamp)
+  VALUES (5, UNIX_TIMESTAMP() * 1000);

schema/mysql/upgrades/optional/1.1.2-history.sql

@@ -0,0 +1 @@
+ALTER TABLE state_history MODIFY COLUMN check_attempt int unsigned NOT NULL;

schema/pgsql/schema.sql

@@ -405,7 +405,7 @@ CREATE TABLE host_state (
   hard_state tinyuint NOT NULL,
   previous_soft_state tinyuint NOT NULL,
   previous_hard_state tinyuint NOT NULL,
-  check_attempt tinyuint NOT NULL,
+  check_attempt uint NOT NULL,
   severity smalluint NOT NULL,
 
   output text DEFAULT NULL,
@@ -675,7 +675,7 @@ CREATE TABLE service_state (
   hard_state tinyuint NOT NULL,
   previous_soft_state tinyuint NOT NULL,
   previous_hard_state tinyuint NOT NULL,
-  check_attempt tinyuint NOT NULL,
+  check_attempt uint NOT NULL,
   severity smalluint NOT NULL,
 
   output text DEFAULT NULL,
@@ -1846,7 +1846,7 @@ CREATE TABLE state_history (
   hard_state tinyuint NOT NULL,
   previous_soft_state tinyuint NOT NULL,
   previous_hard_state tinyuint NOT NULL,
-  check_attempt tinyuint NOT NULL,
+  check_attempt uint NOT NULL, -- may be a tinyuint, see https://icinga.com/docs/icinga-db/latest/doc/04-Upgrading/#upgrading-to-icinga-db-v112
   output text DEFAULT NULL,
   long_output text DEFAULT NULL,
   max_check_attempts uint NOT NULL,
@@ -2181,4 +2181,4 @@ CREATE TABLE icingadb_schema (
 ALTER SEQUENCE icingadb_schema_id_seq OWNED BY icingadb_schema.id;
 
 INSERT INTO icingadb_schema (version, timestamp)
-  VALUES (2, extract(epoch from now()) * 1000);
+  VALUES (3, extract(epoch from now()) * 1000);

schema/pgsql/upgrades/1.1.2.sql

@@ -142,3 +142,12 @@ CREATE INDEX CONCURRENTLY idx_history_event_time_event_type ON history(event_tim
 COMMENT ON INDEX idx_history_event_time_event_type IS 'History filtered/ordered by event_time/event_type';
 
 DROP INDEX idx_history_event_time;
+
+ALTER TABLE host_state ALTER COLUMN check_attempt TYPE uint;
+
+ALTER TABLE service_state ALTER COLUMN check_attempt TYPE uint;
+
+COMMENT ON COLUMN state_history.check_attempt IS 'optional schema upgrade not applied yet, see https://icinga.com/docs/icinga-db/latest/doc/04-Upgrading/#upgrading-to-icinga-db-v112';
+
+INSERT INTO icingadb_schema (version, timestamp)
+  VALUES (3, extract(epoch from now()) * 1000);

schema/pgsql/upgrades/optional/1.1.2-history.sql

@@ -0,0 +1,3 @@
+ALTER TABLE state_history ALTER COLUMN check_attempt TYPE uint;
+
+COMMENT ON COLUMN state_history.check_attempt IS NULL;