Cluster Management
In this section, we discuss how you can add or remove nodes from your LightDB distributed database cluster and how you can deal with node failures.
备注
To make moving shards across nodes or re-replicating shards on failed nodes easier, LightDB distributed database supports fully online shard rebalancing. We discuss briefly the functions provided by the shard rebalancer when relevant in the sections below. You can learn more about these functions, their arguments, and usage, in the Cluster Management And Repair Functions reference section.
Choosing Cluster Size
This section explores configuration settings for running a cluster in production.
Shard Count
Choosing the shard count for each distributed table is a balance between the flexibility of having more shards, and the overhead for query planning and execution across them. If you decide to change the shard count of a table after distributing, you can use the alter_distributed_table function.
Multi-Tenant SaaS Use-Case
The optimal choice varies depending on your access patterns for the data. For instance, use-case we recommend choosing between 32 - 128 shards. For smaller workloads say <100GB, you could start with 32 shards and for larger workloads you could choose 64 or 128. This means that you have the leeway to scale from 32 to 128 worker machines.
Real-Time Analytics Use-Case
Use-case, shard count should be related to the total number of cores on the workers. To ensure maximum parallelism, you should create enough shards on each node such that there is at least one shard per CPU core. We typically recommend creating a high number of initial shards, e.g. 2x or 4x the number of current CPU cores. This allows for future scaling if you add more workers and CPU cores.
However, keep in mind that for each query LightDB distributed opens one database connection per shard, and these connections are limited. Be careful to keep the shard count small enough that distributed queries won’t often have to wait for a connection. Put another way, the connections needed, (max concurrent queries * shard count), should generally not exceed the total connections possible in the system, (number of workers * max_connections per worker).
Initial Hardware Size
The size of a cluster, in terms of number of nodes and their hardware capacity, is easy to change. However, you still need to choose an initial size for a new cluster. Here are some tips for a reasonable initial cluster size.
Multi-Tenant SaaS Use-Case
For those migrating to LightDB distributed from an existing single-node database instance, we recommend choosing a cluster where the number of worker cores and RAM in total equals that of the original instance. In such scenarios we have seen 2-3x performance improvements because sharding improves resource utilization, allowing smaller indices etc.
The coordinator node needs less memory than workers, so you can choose a compute-optimized machine for running the coordinator. The number of cores required depends on your existing workload (write/read throughput).
Real-Time Analytics Use-Case
Total cores: when working data fits in RAM, you can expect a linear performance improvement on LightDB distributed database proportional to the number of worker cores. To determine the right number of cores for your needs, consider the current latency for queries in your single-node database and the required latency in LightDB distributed database. Divide current latency by desired latency, and round the result.
Worker RAM: the best case would be providing enough memory that the majority of the working set fits in memory. The type of queries your application uses affect memory requirements. You can run EXPLAIN ANALYZE on a query to determine how much memory it requires.
Scaling the cluster
Canopy’s logical sharding based architecture allows you to scale out your cluster without any downtime. This section describes how you can add more nodes to your LightDB distributed database cluster in order to improve query performance / scalability.
Add a worker
LightDB distributed database stores all the data for distributed tables on the worker nodes. Hence, if you want to scale out your cluster by adding more computing power, you can do so by adding a worker.
To add a new node to the cluster, you first need to add the DNS name or IP address of that node and port (on which LightDB is running) in the pg_dist_node catalog table. You can do so using the canopy_add_node UDF. Example:
SELECT * from canopy_add_node('node-name', 5432);
The new node is available for shards of new distributed tables. Existing shards will stay where they are unless redistributed, so adding a new worker may not help performance without further steps.
If your cluster has very large reference tables, they can slow down the addition of a node. In this case, consider the canopy.replicate_reference_tables_on_activate (boolean) GUC.
备注
LightDB distributed database workers use encrypted communication by default. A new node will refuse to talk with other workers who do not have SSL enabled. When adding a node to a cluster without encrypted communication, you must reconfigure the new node before creating the LightDB distributed database.
First, from the coordinator node check whether the other workers use SSL:
SELECT run_command_on_workers('show ssl');
If they do not, then connect to the new node and permit it to communicate over plaintext if necessary:
ALTER SYSTEM SET canopy.node_conninfo TO 'sslmode=prefer';
SELECT pg_reload_conf();
Rebalance Shards without Downtime
备注
LightDB distributed database now supports non-blocking reads and writes during rebalancing.
If you want to move existing shards to a newly added worker, LightDB distributed database provides a rebalance_table_shards function to make it easier. This function will move the shards of a given table to distribute them evenly among the workers.
The function is configurable to rebalance shards according to a number of strategies, to best match your database workload. See the function reference to learn which strategy to choose. Here’s an example of rebalancing shards using the default strategy:
SELECT rebalance_table_shards();
Many products, like multi-tenant SaaS applications, cannot tolerate downtime, and on our managed service, rebalancing is able to honor this requirement on Lightdb. This means reads and writes from the application can continue with minimal interruption while data is being moved.
How it Works
LightDB distributed database’s shard rebalancing uses Lightdb logical replication to move data from the old shard (called the “publisher” in replication terms) to the new (the “subscriber.”) Logical replication allows application reads and writes to continue uninterrupted while copying shard data. LightDB distributed database puts a brief write-lock on a shard only during the time it takes to update metadata to promote the subscriber shard as active.
As the Lightdb docs explain, the source needs a replica identity configured:
A published table must have a “replica identity” configured in order to be able to replicate UPDATE and DELETE operations, so that appropriate rows to update or delete can be identified on the subscriber side. By default, this is the primary key, if there is one. Another unique index (with certain additional requirements) can also be set to be the replica identity.
In other words, if your distributed table has a primary key defined then it’s ready for shard rebalancing with no extra work. However, if it doesn’t have a primary key or an explicitly defined replica identity, then attempting to rebalance it will cause an error. For instance:
-- creating the following table without REPLICA IDENTITY or PRIMARY KEY
CREATE TABLE test_table (key int not null, value text not null);
-- add a new worker node to simulate need for
-- shard rebalancing
-- running shard rebalancer with default behavior
SELECT rebalance_table_shards('test_table');
/*
NOTICE: Moving shard 102040 from localhost:9701 to localhost:9700 ...
ERROR: cannot use logical replication to transfer shards of the
relation test_table since it doesn't have a REPLICA IDENTITY or
PRIMARY KEY
DETAIL: UPDATE and DELETE commands on the shard will error out during
logical replication unless there is a REPLICA IDENTIY or PRIMARY KEY.
HINT: If you wish to continue without a replica identity set the
shard_transfer_mode to 'force_logical' or 'block_writes'.
*/
Here’s how to fix this error.
First, does the table have a unique index?
If the table to be replicated already has a unique index which includes the distribution column, then choose that index as a replica identity:
-- supposing my_table has unique index my_table_idx
-- which includes distribution column
ALTER TABLE my_table REPLICA IDENTITY
USING INDEX my_table_idx;
备注
While REPLICA IDENTITY USING INDEX is fine, we recommend against adding REPLICA IDENTITY FULL to a table. This setting would result in each update/delete doing a full-table-scan on the subscriber side to find the tuple with those rows. In our testing we’ve found this to result in worse performance than even solution four below.
Otherwise, can you add a primary key?
Add a primary key to the table. If the desired key happens to be the distribution column, then it’s quite easy, just add the constraint. Otherwise, a primary key with a non-distribution column must be composite and contain the distribution column too.
Unwilling to add primary key or unique index?
If the distributed table doesn’t have a primary key or replica identity, and adding one is unclear or undesirable, you can still force the use of logical replication on LightDB. It’s OK to do this on a table which receives only reads and inserts (no deletes or updates). Include the optional shard_transfer_mode argument of rebalance_table_shards:
SELECT rebalance_table_shards(
'test_table',
shard_transfer_mode => 'force_logical'
);
In this situation if an application does attempt an update or delete during replication, then the request will merely return an error. Deletes and writes will become possible again after replication is complete.
Adding a coordinator
The LightDB distributed database coordinator only stores metadata about the table shards and does not store any data. This means that all the computation is pushed down to the workers and the coordinator does only final aggregations on the result of the workers. Therefore, it is not very likely that the coordinator becomes a bottleneck for read performance. Also, it is easy to boost up the coordinator by shifting to a more powerful machine.
However, in some write heavy use cases where the coordinator becomes a performance bottleneck, users can add another coordinator. As the metadata tables are small (typically a few MBs in size), it is possible to copy over the metadata onto another node and sync it regularly. Once this is done, users can send their queries to any coordinator and scale out performance.
Dealing With Node Failures
In this subsection, we discuss how you can deal with node failures without incurring any downtime on your LightDB distributed database cluster.
Worker Node Failures
LightDB distributed database uses LightDB streaming replication, allowing it to tolerate worker-node failures. This option replicates entire worker nodes by continuously streaming their WAL records to a standby. You can configure streaming replication on-premise yourself by consulting the LightDB replication documentation <https://www.light-pg.com/docs/lightdb/current/warm-standby.html#STREAMING-REPLICATION>>.
Coordinator Node Failures
The LightDB distributed database coordinator maintains metadata tables to track all of the cluster nodes and the locations of the database shards on those nodes. The metadata tables are small (typically a few MBs in size) and do not change very often. This means that they can be replicated and quickly restored if the node ever experiences a failure. There are several options on how users can deal with coordinator failures.
Use LightDB streaming replication: You can use LightDB’s streaming replication feature to create a hot standby of the coordinator. Then, if the primary coordinator node fails, the standby can be promoted to the primary automatically to serve queries to your cluster.
Use backup tools: Since the metadata tables are small, users can use EBS volumes, or LightDB backup tools to backup the metadata. Then, they can easily copy over that metadata to new nodes to resume operation.
Tenant Isolation
备注
LightDB distributed database includes tenant isolation functionality now!
LightDB distributed database places table rows into worker shards based on the hashed value of the rows’ distribution column. Multiple distribution column values often fall into the same shard. In the LightDB distributed database multi-tenant use case this means that tenants often share shards.
However, sharing shards can cause resource contention when tenants differ drastically in size. This is a common situation for systems with a large number of tenants – we have observed that the size of tenant data tend to follow a Zipfian distribution as the number of tenants increases. This means there are a few very large tenants, and many smaller ones. To improve resource allocation and make guarantees of tenant QoS it is worthwhile to move large tenants to dedicated nodes.
LightDB distributed database provides the tools to isolate a tenant on a specific node. This happens in two phases: 1) isolating the tenant’s data to a new dedicated shard, then 2) moving the shard to the desired node. To understand the process it helps to know precisely how rows of data are assigned to shards.
Every shard is marked in LightDB distributed database metadata with the range of hashed values it contains (more info in the reference for pg_dist_shard).
Viewing Query Statistics
备注
LightDB distributed database now includes the canopy_stat_statements view!
When administering a LightDB distributed database cluster it’s useful to know what queries users are running, which nodes are involved, and which execution method LightDB distributed database is using for each query. LightDB distributed database records query statistics in a metadata view called canopy_stat_statements, named analogously to LightDB’ lt_stat_statments. Whereas pg_stat_statements stores info about query duration and I/O, canopy_stat_statements stores info about LightDB distributed database execution methods and shard partition keys (when applicable).
LightDB distributed database requires the lt_stat_statements extension to be installed in order to track query statistics. this extension will be pre-activated in LightDB
Let’s see how this works. Assume we have a table called foo that is hash-distributed by its id column.
-- create and populate distributed table
create table foo ( id int );
insert into foo select generate_series(1,100);
We’ll run two more queries, and canopy_stat_statements will show how LightDB distributed database chooses to execute them.
-- counting all rows executes on all nodes, and sums
-- the results on the coordinator
SELECT count(*) FROM foo;
-- specifying a row by the distribution column routes
-- execution to an individual node
SELECT * FROM foo WHERE id = 42;
To find how these queries were executed, ask the stats table:
SELECT * FROM canopy_stat_statements;
Results:
-[ RECORD 1 ]-+----------------------------------------------
queryid | -6844578505338488014
userid | 10
dbid | 13340
query | SELECT count(*) FROM foo;
executor | adaptive
partition_key |
calls | 1
-[ RECORD 2 ]-+----------------------------------------------
queryid | 185453597994293667
userid | 10
dbid | 13340
query | insert into foo select generate_series($1,$2)
executor | insert-select
partition_key |
calls | 1
-[ RECORD 3 ]-+----------------------------------------------
queryid | 1301170733886649828
userid | 10
dbid | 13340
query | SELECT * FROM foo WHERE id = $1
executor | adaptive
partition_key | 42
calls | 1
We can see that LightDB distributed database uses the adaptive executor most commonly to run queries. This executor fragments the query into constituent queries to run on relevant nodes, and combines the results on the coordinator node. In the case of the second query (filtering by the distribution column id = $1), LightDB distributed database determined that it needed the data from just one node. Lastly, we can see that the insert into foo select… statement ran with the insert-select executor which provides flexibility to run these kind of queries.
So far the information in this view doesn’t give us anything we couldn’t already learn by running the EXPLAIN command for a given query. However, in addition to getting information about individual queries, the canopy_stat_statements view allows us to answer questions such as “what percentage of queries in the cluster are scoped to a single tenant?”
SELECT sum(calls),
partition_key IS NOT NULL AS single_tenant
FROM canopy_stat_statements
GROUP BY 2;
.
sum | single_tenant
-----+---------------
2 | f
1 | t
In a multi-tenant database, for instance, we would expect the vast majority of queries to be single tenant. Seeing too many multi-tenant queries may indicate that queries do not have the proper filters to match a tenant, and are using unnecessary resources.
We can also find which partition_ids are the most frequent targets. In a multi-tenant application these would be the busiest tenants.
SELECT partition_key, sum(calls) as total_queries
FROM canopy_stat_statements
WHERE coalesce(partition_key, '') <> ''
GROUP BY partition_key
ORDER BY total_queries desc
LIMIT 10;
┌───────────────┬───────────────┐
│ partition_key │ total_queries │
├───────────────┼───────────────┤
│ 42 │ 1 │
└───────────────┴───────────────┘
Statistics Expiration
The lt_stat_statements view limits the number of statements it tracks, and the duration of its records. Because canopy_stat_statements tracks a strict subset of the queries in lt_stat_statements, a choice of equal limits for the two views would cause a mismatch in their data retention. Mismatched records can cause joins between the views to behave unpredictably.
There are three ways to help synchronize the views, and all three can be used together.
Have the maintenance daemon periodically sync the LightDB distributed database and lt stats. The GUC
canopy.stat_statements_purge_intervalsets time in seconds for the sync. A value of 0 disables periodic syncs.Adjust the number of entries in canopy_stat_statements. The
canopy.stat_statements_maxGUC removes old entries when new ones cross the threshold. The default value is 50K, and the highest allowable value is 10M. Note that each entry costs about 140 bytes in shared memory so set the value wisely.Increase
lt_stat_statements.max. Its default value is 5000, and could be increased to 10K, 20K or even 50K without much overhead. This is most beneficial when there is more local (i.e. coordinator) query workload.
备注
Changing lt_stat_statements.max or canopy.stat_statements_max requires restarting the LightDB service. Changing canopy.stat_statements_purge_interval, on the other hand, will come into effect with a call to pg_reload_conf().
Resource Conservation
Limiting Long-Running Queries
Long running queries can hold locks, queue up WAL, or just consume a lot of system resources, so in a production environment it’s good to prevent them from running too long. You can set the statement_timeout parameter on the coordinator and workers to cancel queries that run too long.
-- limit queries to five minutes
ALTER DATABASE distributed
SET statement_timeout TO 300000;
SELECT run_command_on_workers($cmd$
ALTER DATABASE distributed
SET statement_timeout TO 300000;
$cmd$);
The timeout is specified in milliseconds.
To customize the timeout per query, use SET LOCAL in a transaction:
BEGIN;
-- this limit applies to just the current transaction
SET LOCAL statement_timeout TO 300000;
-- ...
COMMIT;
Security
Connection Management
备注
The traffic between the different nodes in the cluster is encrypted for NEW installations. This is done by using TLS with self-signed certificates. This means that this does not protect against Man-In-The-Middle attacks. This only protects against passive eavesdropping on the network.
When LightDB distributed database nodes communicate with one another they consult a table with connection credentials. This gives the database administrator flexibility to adjust parameters for security and efficiency.
To set non-sensitive libpq connection parameters to be used for all node connections, update the canopy.node_conninfo GUC:
-- key=value pairs separated by spaces.
-- For example, ssl options:
ALTER SYSTEM SET canopy.node_conninfo =
'sslrootcert=/path/to/canopy-ca.crt sslcrl=/path/to/canopy-ca.crl sslmode=verify-full';
There is a whitelist of parameters that the GUC accepts, see the node_conninfo reference for details. The default value for node_conninfo is sslmode=require, which prevents unencrypted communication between nodes.
After changing this setting it is important to reload the lightdb configuration. Even though the changed setting might be visible in all sessions, the setting is only consulted by LightDB distributed database when new connections are established. When a reload signal is received, LightDB distributed database marks all existing connections to be closed which causes a reconnect after running transactions have been completed.
SELECT pg_reload_conf();
-- only superusers can access this table
-- add a password for user jdoe
INSERT INTO pg_dist_authinfo
(nodeid, rolename, authinfo)
VALUES
(123, 'jdoe', 'password=abc123');
After this INSERT, any query needing to connect to node 123 as the user jdoe will use the supplied password. The documentation for pg_dist_authinfo has more info.
-- update user jdoe to use certificate authentication
UPDATE pg_dist_authinfo
SET authinfo = 'sslcert=/path/to/user.crt sslkey=/path/to/user.key'
WHERE nodeid = 123 AND rolename = 'jdoe';
This changes the user from using a password to use a certificate and keyfile while connecting to node 123 instead. Make sure the user certificate is signed by a certificate that is trusted by the worker you are connecting to and authentication settings on the worker allow for certificate based authentication. Full documentation on how to use client certificates can be found in the lightdb libpq documentation.
Changing pg_dist_authinfo does not force any existing connection to reconnect.
Setup Certificate Authority signed certificates
This section assumes you have a trusted Certificate Authority that can issue server certificates to you for all nodes in your cluster. It is recommended to work with the security department in your organization to prevent key material from being handled incorrectly. This guide covers only LightDB distributed database specific configuration that needs to be applied, not best practices for PKI management.
For all nodes in the cluster you need to get a valid certificate signed by the same Certificate Authority. The following machine specific files are assumed to be available on every machine:
/path/to/server.key: Server Private Key/path/to/server.crt: Server Certificate or Certificate Chain for Server Key, signed by trusted Certificate Authority.
Next to these machine specific files you need these cluster or CA wide files available:
/path/to/ca.crt: Certificate of the Certificate Authority/path/to/ca.crl: Certificate Revocation List of the Certificate Authority
备注
The Certificate Revocation List is likely to change over time. Work with your security department to set up a mechanism to update the revocation list on to all nodes in the cluster in a timely manner. A reload of every node in the cluster is required after the revocation list has been updated.
Once all files are in place on the nodes, the following settings need to be configured in the LightDB configuration file:
# the following settings allow the lightdb server to enable ssl, and
# configure the server to present the certificate to clients when
# connecting over tls/ssl
ssl = on
ssl_key_file = '/path/to/server.key'
ssl_cert_file = '/path/to/server.crt'
# this will tell LightDB distributed database to verify the certificate of the server it is connecting to
canopy.node_conninfo = 'sslmode=verify-full sslrootcert=/path/to/ca.crt sslcrl=/path/to/ca.crl'
After changing, either restart the database or reload the configuration to apply these changes. Also, adjusting canopy.local_hostname (text) may be required for proper functioning with sslmode=verify-full.
Depending on the policy of the Certificate Authority used you might need or want to change sslmode=verify-full in canopy.node_conninfo to sslmode=verify-ca. For the difference between the two settings please consult the official lightdb documentation.
Lastly, to prevent any user from connecting via an un-encrypted connection, changes need to be made to lt_hba.conf. Many LightDB installations will have entries allowing host connections which allow SSL/TLS connections as well as plain TCP connections. By replacing all host entries with hostssl entries, only encrypted connections will be allowed to authenticate to LightDB. For full documentation on these settings take a look at the lt_hba.conf file documentation on the official LightDB documentation.
备注
When a trusted Certificate Authority is not available, one can create their own via a self-signed root certificate. This is non-trivial and the developer or operator should seek guidance from their security team when doing so.
To verify the connections from the coordinator to the workers are encrypted you can run the following query. It will show the SSL/TLS version used to encrypt the connection that the coordinator uses to talk to the worker:
SELECT run_command_on_workers($$
SELECT version FROM pg_stat_ssl WHERE pid = pg_backend_pid()
$$);
┌────────────────────────────┐
│ run_command_on_workers │
├────────────────────────────┤
│ (localhost,9701,t,TLSv1.2) │
│ (localhost,9702,t,TLSv1.2) │
└────────────────────────────┘
(2 rows)
Increasing Worker Security
For your convenience getting started, our multi-node installation instructions direct you to set up the lt_hba.conf on the workers with its authentication method set to “trust” for local network connections. However, you might desire more security.
To require that all connections supply a hashed password, update the LightDB lt_hba.conf on every worker node with something like this:
# Require password access and a ssl/tls connection to nodes in the local
# network. The following ranges correspond to 24, 20, and 16-bit blocks
# in Private IPv4 address spaces.
hostssl all all 10.0.0.0/8 md5
# Require passwords and ssl/tls connections when the host connects to
# itself as well.
hostssl all all 127.0.0.1/32 md5
hostssl all all ::1/128 md5
The coordinator node needs to know roles’ passwords in order to communicate with the workers. In LightDB distributed database the authentication information has to be maintained in a .pgpass file. Edit .pgpass in the lightdb user’s home directory, with a line for each combination of worker address and role:
hostname:port:database:username:password
Sometimes workers need to connect to one another, such as during repartition joins. Thus each worker node requires a copy of the .pgpass file as well.
Row-Level Security
备注
LightDB distributed database now supports row-level security for distributed tables.
LightDB row-level security policies restrict, on a per-user basis, which rows can be returned by normal queries or inserted, updated, or deleted by data modification commands. This can be especially useful in a multi-tenant LightDB distributed database cluster because it allows individual tenants to have full SQL access to the database while hiding each tenant’s information from other tenants.
We can implement the separation of tenant data by using a naming convention for database roles that ties into table row-level security policies. We’ll assign each tenant a database role in a numbered sequence: tenant_1, tenant_2, etc. Tenants will connect to LightDB distributed database using these separate roles. Row-level security policies can compare the role name to values in the tenant_id distribution column to decide whether to allow access.
Here is how to apply the approach on a simplified events table distributed by tenant_id. First create the roles tenant_1 and tenant_2. Then run the following as an administrator:
CREATE TABLE events(
tenant_id int,
id int,
type text
);
INSERT INTO events VALUES (1,1,'foo'), (2,2,'bar');
-- assumes that roles tenant_1 and tenant_2 exist
GRANT select, update, insert, delete
ON events TO tenant_1, tenant_2;
As it stands, anyone with select permissions for this table can see both rows. Users from either tenant can see and update the row of the other tenant. We can solve this with row-level table security policies.
Each policy consists of two clauses: USING and WITH CHECK. When a user tries to read or write rows, the database evaluates each row against these clauses. Existing table rows are checked against the expression specified in USING, while new rows that would be created via INSERT or UPDATE are checked against the expression specified in WITH CHECK.
-- first a policy for the system admin "distributed" user
CREATE POLICY admin_all ON events
TO distributed -- apply to this role
USING (true) -- read any existing row
WITH CHECK (true); -- insert or update any row
-- next a policy which allows role "tenant_<n>" to
-- access rows where tenant_id = <n>
CREATE POLICY user_mod ON events
USING (current_user = 'tenant_' || tenant_id::text);
-- lack of CHECK means same condition as USING
-- enforce the policies
ALTER TABLE events ENABLE ROW LEVEL SECURITY;
Now roles tenant_1 and tenant_2 get different results for their queries:
Connected as tenant_1:
SELECT * FROM events;
┌───────────┬────┬──────┐
│ tenant_id │ id │ type │
├───────────┼────┼──────┤
│ 1 │ 1 │ foo │
└───────────┴────┴──────┘
Connected as tenant_2:
SELECT * FROM events;
┌───────────┬────┬──────┐
│ tenant_id │ id │ type │
├───────────┼────┼──────┤
│ 2 │ 2 │ bar │
└───────────┴────┴──────┘
INSERT INTO events VALUES (3,3,'surprise');
/*
ERROR: new row violates row-level security policy for table "events_102055"
*/
LightDB extensions
LightDB distributed database provides distributed functionality by extending LightDB using the hook and extension APIs. This allows users to benefit from the features that come with the rich LightDB ecosystem. These features include, but aren’t limited to, support for a wide range of data types (including semi-structured data types like jsonb and hstore), operators and functions, full text search, and other extensions such as PostGIS. Further, proper use of the extension APIs enable compatibility with standard LightDB tools such as pgAdmin and lt_upgrade.
As LightDB distributed database which can be installed on any LightDB instance, you can directly use other extensions such as hstore, or PostGIS with LightDB distributed database. However, there is one thing to keep in mind. While including other extensions in shared_preload_libraries, you should make sure that LightDB distributed is the first extension.
Creating a New Database
Each LightDB server can hold multiple databases. However, new databases do not inherit the extensions of any others; all desired extensions must be added afresh. To run LightDB distributed on a new database, you’ll need to create the database on the coordinator and workers, create the LightDB distributed within that database, and register the workers in the coordinator database.
Connect to each of the worker nodes and run:
-- on every worker node
CREATE DATABASE newbie;
\c newbie
Then, on the coordinator:
CREATE DATABASE newbie;
\c newbie
SELECT * from canopy_add_node('node-name', 5432);
SELECT * from canopy_add_node('node-name2', 5432);
-- ... for all of them
Now the new database will be operating as another LightDB distributed database cluster.