A more in-depth discussion of various gh-ost command line flags: implementation, implication, use cases.
Add this flag when executing on Aliyun RDS.
Allows the user to make schema changes that include a zero date or zero in date (e.g. adding a datetime default '0000-00-00 00:00:00' column), even if global sql_mode on MySQL has NO_ZERO_IN_DATE,NO_ZERO_DATE.
Add this flag when executing on Azure Database for MySQL.
See --assume-master-host.
By default, gh-ost would like you to connect to a replica, from where it figures out the master by itself. This wiring is required should your master execute using binlog_format=STATEMENT.
If, for some reason, you do not wish gh-ost to connect to a replica, you may connect it directly to the master and approve this via --allow-on-master.
When your migration issues a column rename (change column old_name new_name ...) gh-ost analyzes the statement to try and associate the old column name with new column name. Otherwise, the new structure may also look like some column was dropped and another was added.
gh-ost will print out what it thinks the rename implied, but will not issue the migration unless you provide with --approve-renamed-columns.
If you think gh-ost is mistaken and that there's actually no rename involved, you may pass --skip-renamed-columns instead. This will cause gh-ost to disassociate the column values; data will not be copied between those columns.
gh-ost infers the identity of the master server by crawling up the replication topology. You may explicitly tell gh-ost the identity of the master host via --assume-master-host=the.master.com. This is useful in:
- master-master topologies (together with
--allow-master-master), wheregh-ostcan arbitrarily pick one of the co-masters, and you prefer that it picks a specific one - tungsten replicator topologies (together with
--tungsten), wheregh-ostis unable to crawl and detect the master
If you happen to know your servers use RBR (Row Based Replication, i.e. binlog_format=ROW), you may specify --assume-rbr. This skips a verification step where gh-ost would issue a STOP SLAVE; START SLAVE.
Skipping this step means gh-ost would not need the SUPER privilege in order to operate.
You may want to use this on Amazon RDS.
MySQL 8.0 supports "instant DDL" for some operations. If an alter statement can be completed with instant DDL, only a metadata change is required internally. Instant operations include:
- Adding a column
- Dropping a column
- Dropping an index
- Extending a varchar column
- Adding a virtual generated column
It is not reliable to parse the ALTER statement to determine if it is instant or not. This is because the table might be in an older row format, or have some other incompatibility that is difficult to identify.
--attempt-instant-ddl is disabled by default, but the risks of enabling it are relatively minor: gh-ost may need to acquire a metadata lock at the start of the operation. This is not a problem for most scenarios, but it could be a problem for users that start the DDL during a period with long running transactions.
gh-ost will automatically fallback to the normal DDL process if the attempt to use instant DDL is unsuccessful.
--binlogsyncer-max-reconnect-attempts=0, the maximum number of attempts to re-establish a broken inspector connection for sync binlog. 0 or negative number means infinite retry, default 0
--conf=/path/to/my.cnf: file where credentials are specified. Should be in (or contain) the following format:
[client]
user=gromit
password=123456
Defaults to true. See exact-rowcount
Comma delimited status-name=threshold, same format as --max-load.
--critical-load defines a threshold that, when met, gh-ost panics and bails out. The default behavior is to bail out immediately when meeting this threshold.
This may sometimes lead to migrations bailing out on a very short spike, that, while in itself is impacting production and is worth investigating, isn't reason enough to kill a 10-hour migration.
When --critical-load-hibernate-seconds is non-zero (e.g. --critical-load-hibernate-seconds=300), critical-load does not panic and bail out; instead, gh-ost goes into hibernation for the specified duration. It will not read/write anything from/to any server during this time. Execution then continues upon waking from hibernation.
If critical-load is met again, gh-ost will repeat this cycle, and never panic and bail out.
When --critical-load-interval-millis is specified (e.g. --critical-load-interval-millis=2500), gh-ost gives a second chance: when it meets critical-load threshold, it doesn't bail out. Instead, it starts a timer (in this example: 2.5 seconds) and re-checks critical-load when the timer expires. If critical-load is met again, gh-ost panics and bails out. If not, execution continues.
This is somewhat similar to a Nagios n-times test, where n in our case is always 2.
Optional. Default is safe. See more discussion in cut-over
Default 3. Max number of seconds to hold locks on tables while attempting to cut-over (retry attempted when lock exceeds timeout).
Danger: this flag will silently discard any foreign keys existing on your table.
At this time (10-2016) gh-ost does not support foreign keys on migrated tables (it bails out when it notices a FK on the migrated table). However, it is able to support dropping of foreign keys via this flag. If you're trying to get rid of foreign keys in your environment, this is a useful flag.
See also: skip-foreign-key-checks
gh-ost reads event from the binary log and applies them onto the ghost table. It does so in batched writes: grouping multiple events to apply in a single transaction. This gives better write throughput as we don't need to sync the transaction log to disk for each event.
The --dml-batch-size flag controls the size of the batched write. Allowed values are 1 - 100, where 1 means no batching (every event from the binary log is applied onto the ghost table on its own transaction). Default value is 10.
Why is this behavior configurable? Different workloads have different characteristics. Some workloads have very large writes, such that aggregating even 50 writes into a transaction makes for a significant transaction size. On other workloads write rate is high such that one just can't allow for a hundred more syncs to disk per second. The default value of 10 is a modest compromise that should probably work very well for most workloads. Your mileage may vary.
Noteworthy is that setting --dml-batch-size to higher value does not mean gh-ost blocks or waits on writes. The batch size is an upper limit on transaction size, not a minimal one. If gh-ost doesn't have "enough" events in the pipe, it does not wait on the binary log, it just writes what it already has. This conveniently suggests that if write load is light enough for gh-ost to only see a few events in the binary log at a given time, then it is also light enough for gh-ost to apply a fraction of the batch size.
A gh-ost execution need to copy whatever rows you have in your existing table onto the ghost table. This can and often will be, a large number. Exactly what that number is?
gh-ost initially estimates the number of rows in your table by issuing an explain select * from your_table. This will use statistics on your table and return with a rough estimate. How rough? It might go as low as half or as high as double the actual number of rows in your table. This is the same method as used in pt-online-schema-change.
gh-ost also supports the --exact-rowcount flag. When this flag is given, two things happen:
- An initial, authoritative
select count(*) from your_table. This query may take a long time to complete, but is performed before we begin the massive operations. When--concurrent-rowcountis also specified, this runs in parallel to row copy. Note:--concurrent-rowcountnow defaults totrue. - A continuous update to the estimate as we make progress applying events. We heuristically update the number of rows based on the queries we process from the binlogs.
While the ongoing estimated number of rows is still heuristic, it's almost exact, such that the reported ETA or percentage progress is typically accurate to the second throughout a multiple-hour operation.
Without this parameter, migration is a noop: testing table creation and validity of migration, but not touching data.
If given, a cut-over command must name the migrated table, or else ignored.
If given, a panic command must name the migrated table, or else ignored.
Table name prefix to be used on the temporary tables.
Add this flag when executing on a 1st generation Google Cloud Platform (GCP).
Default 100. See subsecond-lag for details.
Defaults to 60 seconds. Configures how often the gh-ost-on-status hook is called, see hooks for full details on how to use hooks.
gh-ost maintains two tables while migrating: the ghost table (which is synced from your original table and finally replaces it) and a changelog table, which is used internally for bookkeeping. By default, it panics and aborts if it sees those tables upon startup. Provide --initially-drop-ghost-table and --initially-drop-old-table to let gh-ost know it's OK to drop them beforehand.
We think gh-ost should not take chances or make assumptions about the user's tables. Dropping tables can be a dangerous, locking operation. We let the user explicitly approve such operations.
See initially-drop-ghost-table
Default False. Should gh-ost forcibly delete an existing socket file. Be careful: this might drop the socket file of a running migration!
On a replication topology, this is perhaps the most important migration throttling factor: the maximum lag allowed for migration to work. If lag exceeds this value, migration throttles.
When using Connect to replica, migrate on master, this lag is primarily tested on the very replica gh-ost operates on. Lag is measured by checking the heartbeat events injected by gh-ost itself on the utility changelog table. That is, to measure this replica's lag, gh-ost doesn't need to issue show slave status nor have any external heartbeat mechanism.
When --throttle-control-replicas is provided, throttling also considers lag on specified hosts. Lag measurements on listed hosts is done by querying gh-ost's changelog table, where gh-ost injects a heartbeat.
When using on master or when --allow-on-master is provided, max-lag-millis is also considered a threshold for starting the cutover stage of the migration. If the row copy is complete and the heartbeat lag is less than max-lag-millis cutover phase of the migration will start.
See also: Sub-second replication lag throttling
List of metrics and threshold values; topping the threshold of any will cause throttler to kick in. See also: throttling
Typically gh-ost is used to migrate tables on a master. If you wish to only perform the migration in full on a replica, connect gh-ost to said replica and pass --migrate-on-replica. gh-ost will briefly connect to the master but otherwise will make no changes on the master. Migration will be fully executed on the replica, while making sure to maintain a small replication lag.
Indicate a file name, such that the final cut-over step does not take place as long as the file exists.
When this flag is set, gh-ost expects the file to exist on startup, or else tries to create it. gh-ost exits with error if the file does not exist and gh-ost is unable to create it.
With this flag set, the migration will cut-over upon deletion of the file or upon cut-over interactive command.
Defaults to 99999. If you run multiple migrations then you must provide a different, unique --replica-server-id for each gh-ost process.
Optionally involve the process ID, for example: --replica-server-id=$((1000000000+$$)).
It's on you to choose a number that does not collide with another gh-ost or another running replica.
See also: concurrent-migrations on the cheatsheet.
Defaults to an auto-determined and advertised upon startup file. Defines Unix socket file to serve on.
By default gh-ost verifies no foreign keys exist on the migrated table. On servers with large number of tables this check can take a long time. If you're absolutely certain no foreign keys exist (table does not reference other table nor is referenced by other tables) and wish to save the check time, provide with --skip-foreign-key-checks.
By default gh-ost enforces STRICT_ALL_TABLES sql_mode as a safety measure. In some cases this changes the behaviour of other modes (namely ERROR_FOR_DIVISION_BY_ZERO, NO_ZERO_DATE, and NO_ZERO_IN_DATE) which may lead to errors during migration. Use --skip-strict-mode to explicitly tell gh-ost not to enforce this. Danger This may have some unexpected disastrous side effects.
By default gh-ost does not use ssl/tls connections to the database servers when performing migrations. This flag instructs gh-ost to use encrypted connections. If enabled, gh-ost will use the system's ca certificate pool for server certificate verification. If a different certificate is needed for server verification, see --ssl-ca. If you wish to skip server verification, but still use encrypted connections, use with --ssl-allow-insecure.
Allows gh-ost to connect to the MySQL servers using encrypted connections, but without verifying the validity of the certificate provided by the server during the connection. Requires --ssl.
--ssl-ca=/path/to/ca-cert.pem: ca certificate file (in PEM format) to use for server certificate verification. If specified, the default system ca cert pool will not be used for verification, only the ca cert provided here. Requires --ssl.
--ssl-cert=/path/to/ssl-cert.crt: SSL public key certificate file (in PEM format).
--ssl-key=/path/to/ssl-key.key: SSL private key file (in PEM format).
Default is innodb, and rocksdb support is currently experimental. InnoDB and RocksDB are both transactional engines, supporting both shared and exclusive row locks.
But RocksDB currently lacks a few features support compared to InnoDB:
- Gap Locks
- Foreign Key
- Generated Columns
- Spatial
- Geometry
When --storage-engine=rocksdb, gh-ost will make some changes necessary (e.g. sets isolation level to READ_COMMITTED) to support RocksDB.
The default charset for the database connection is utf8mb4, utf8, latin1. The ability to specify character set and collation is supported, eg: utf8mb4_general_ci,utf8_general_ci,latin1.
Issue the migration on a replica; do not modify data on master. Useful for validating, testing and benchmarking. See testing-on-replica
Default False. When --test-on-replica is enabled, do not issue commands stop replication (requires --test-on-replica).
Provide a command delimited list of replicas; gh-ost will throttle when any of the given replicas lag beyond --max-lag-millis. The list can be queried and updated dynamically via interactive commands
Provide an HTTP endpoint; gh-ost will issue HEAD requests on given URL and throttle whenever response status code is not 200. The URL can be queried and updated dynamically via interactive commands. Empty URL disables the HTTP check.
Defaults to 100. Configures the HTTP throttle check interval in milliseconds.
Defaults to 1000 (1 second). Configures the HTTP throttler check timeout in milliseconds.
Makes the old table include a timestamp value. The old table is what the original table is renamed to at the end of a successful migration. For example, if the table is gh_ost_test, then the old table would normally be _gh_ost_test_del. With --timestamp-old-table it would be, for example, _gh_ost_test_20170221103147_del.
See tungsten on the cheatsheet.