Adds mirror segments to a Greenplum Database system that was initially configured without mirroring.
gpaddmirrors [-p <port_offset>] [-m <datadir_config_file> [-a]] [-s] [-d <master_data_directory>] [-b <segment_batch_size>] [-B <batch_size>] [-l <logfile_directory>] [-v] [--hba-hostnames <boolean>] gpaddmirrors -i <mirror_config_file> [-a] [-d <master_data_directory>] [-b <segment_batch_size>] [-B <batch_size>] [-l <logfile_directory>] [-v] gpaddmirrors -o output_sample_mirror_config> [-s] [-m <datadir_config_file>] gpaddmirrors -? gpaddmirrors --version
gpaddmirrors utility configures mirror segment instances for an existing Greenplum Database system that was initially configured with primary segment instances only. The utility will create the mirror instances and begin the online replication process between the primary and mirror segment instances. Once all mirrors are synchronized with their primaries, your Greenplum Database system is fully data redundant.
Important: During the online replication process, Greenplum Database should be in a quiescent state, workloads and other queries should not be running.
By default, the utility will prompt you for the file system location(s) where it will create the mirror segment data directories. If you do not want to be prompted, you can pass in a file containing the file system locations using the
The mirror locations and ports must be different than your primary segment data locations and ports.
The utility creates a unique data directory for each mirror segment instance in the specified location using the predefined naming convention. There must be the same number of file system locations declared for mirror segment instances as for primary segment instances. It is OK to specify the same directory name multiple times if you want your mirror data directories created in the same location, or you can enter a different data location for each mirror. Enter the absolute path. For example:
Enter mirror segment data directory location 1 of 2 > /gpdb/mirror Enter mirror segment data directory location 2 of 2 > /gpdb/mirror
Enter mirror segment data directory location 1 of 2 > /gpdb/m1 Enter mirror segment data directory location 2 of 2 > /gpdb/m2
Alternatively, you can run the
gpaddmirrors utility and supply a detailed configuration file using the
-i option. This is useful if you want your mirror segments on a completely different set of hosts than your primary segments. The format of the mirror configuration file is:
<contentID> is the segment instance content ID,
<address> is the host name or IP address of the segment host,
<port> is the communication port, and
<data_dir> is the segment instance data directory.
gp_segment_configuration system catalog table can help you determine your current primary segment configuration so that you can plan your mirror segment configuration. For example, run the following query:
=# SELECT dbid, content, address as host_address, port, datadir FROM gp_segment_configuration ORDER BY dbid;
If you are creating mirrors on alternate mirror hosts, the new mirror segment hosts must be pre-installed with the Greenplum Database software and configured exactly the same as the existing primary segment hosts.
You must make sure that the user who runs
gpadmin user) has permissions to write to the data directory locations specified. You may want to create these directories on the segment hosts and
chown them to the appropriate user before running
Note: This utility uses secure shell (SSH) connections between systems to perform its tasks. In large Greenplum Database deployments, cloud deployments, or deployments with a large number of segments per host, this utility may exceed the host's maximum threshold for unauthenticated connections. Consider updating the SSH
MaxStartups configuration parameter to increase this threshold. For more information about SSH configuration options, refer to the SSH documentation for your Linux distribution.
-iif this option is used.
128. If not specified, the utility will start recovering up to 64 segments in parallel on each host.
$MASTER_DATA_DIRECTORYwill be used.
pg_hba.conffile when updating this file with addresses that can connect to Greenplum Database. When set to 0 -- the default value -- this utility uses IP addresses when updating this file. When set to 1, this utility uses host names when updating this file. For consistency, use the same value that was specified for
HBA_HOSTNAMESwhen the Greenplum Database system was initialized. For information about how Greenplum Database resolves host names in the
pg_hba.conffile, see Configuring Client Authentication.
A configuration file containing one line for each mirror segment you want to create. You must have one mirror segment instance listed for each primary segment in the system. The format of this file is as follows (as per attributes in the gp_segment_configuration catalog table):
<contentID> is the segment instance content ID,
<address> is the hostname or IP address of the segment host,
<port> is the communication port, and
<data_dir> is the segment instance data directory. For information about using a hostname or IP address, see Specifying Hosts using Hostnames or IP Addresses. Also, see Using Host Systems with Multiple NICs.
A configuration file containing a list of file system locations where the mirror data directories will be created. If not supplied, the utility prompts you for locations. Each line in the file specifies a mirror data directory location. For example:
/gpdata/m1 /gpdata/m2 /gpdata/m3 /gpdata/m4
-ioption, you can run
gpaddmirrorswith this option to generate a sample mirror configuration file based on your primary segment configuration. The utility will prompt you for your mirror segment data directory locations (unless you provide these in a file using
-m). You can then edit this file to change the host names to alternate mirror hosts if necessary.
Optional. This number is used to calculate the database ports used for mirror segments. The default offset is 1000. Mirror port assignments are calculated as follows:
primary_port + offset = mirror_database_port
When specifying a mirroring configuration using the
-i, you can specify either a hostname or an IP address for the <address> value.
/etc/hostsfile to map the hostname to an IP address. The resolution of a hostname to an IP address should not be performed by an external service such as a public DNS server. You must stop the Greenplum system before you change the mapping of a hostname to a different IP address.
-ioption unless you have a specific requirement to use IP addresses.
When enabling a mirroring configuration that adds hosts to the Greenplum system,
gpaddmirrors populates the gp_segment_configuration catalog table with the mirror segment instance information. Greenplum Database uses the address value of the
gp_segment_configuration catalog table when looking up host systems for Greenplum interconnect (internal) communication between the master and segment instances and between segment instances, and for other internal communication.
If hosts systems are configured with multiple NICs, you can initialize a Greenplum Database system to use each NIC as a Greenplum host system. You must ensure that the host systems are configured with sufficient resources to support all the segment instances being added to the host. Also, if you enable segment mirroring, you must ensure that the Greenplum system configuration supports failover if a host system fails. For information about Greenplum Database mirroring schemes, see Segment Mirroring Configurations.
For example, this is a segment instance configuration for a simple Greenplum system. The segment host
gp6m is configured with two NICs,
gp6m-2, where the Greenplum Database system uses
gp6m-1 for the master segment and
gp6m-2 for segment instances.
select content, role, port, hostname, address from gp_segment_configuration ; content | role | port | hostname | address ---------+------+-------+----------+---------- -1 | p | 5432 | gp6m | gp6m-1 0 | p | 40000 | gp6m | gp6m-2 0 | m | 50000 | gp6s | gp6s 1 | p | 40000 | gp6s | gp6s 1 | m | 50000 | gp6m | gp6m-2 (5 rows)
Add mirroring to an existing Greenplum Database system using the same set of hosts as your primary data. Calculate the mirror database ports by adding 100 to the current primary segment port numbers:
$ gpaddmirrors -p 100
Generate a sample mirror configuration file with the
-o option to use with
$ gpaddmirrors -o /home/gpadmin/sample_mirror_config
Add mirroring to an existing Greenplum Database system using a different set of hosts from your primary data:
$ gpaddmirrors -i mirror_config_file
mirror_config_file looks something like this:
0|sdw1-1|52001|/gpdata/m1/gp0 1|sdw1-2|52002|/gpdata/m2/gp1 2|sdw2-1|52001|/gpdata/m1/gp2 3|sdw2-2|52002|/gpdata/m2/gp3