Canopy Tables and Views
Coordinator Metadata
Canopy divides each distributed table into multiple logical shards based on the distribution column. The coordinator then maintains metadata tables to track statistics and information about the health and location of these shards. In this section, we describe each of these metadata tables and their schema. You can view and query these tables using SQL after logging into the coordinator node.
Partition table
The pg_dist_partition table stores metadata about which tables in the database are distributed. For each distributed table, it also stores information about the distribution method and detailed information about the distribution column.
Name |
Type |
Description |
|---|---|---|
logicalrelid |
regclass |
Distributed table to which this row corresponds. This value references
the relfilenode column in the pg_class system catalog table.
|
partmethod |
char |
The method used for partitioning / distribution. The values of this
column corresponding to different distribution methods are :-
append: ‘a’
hash: ‘h’
reference table: ‘n’
|
partkey |
text |
Detailed information about the distribution column including column
number, type and other relevant information.
|
colocationid |
integer |
Co-location group to which this table belongs. Tables in the same group
allow co-located joins and distributed rollups among other
optimizations. This value references the colocationid column in the
pg_dist_colocation table.
|
repmodel |
char |
The method used for data replication. The values of this column
corresponding to different replication methods are :-
* canopy statement-based replication: ‘c’
* lightdb streaming replication: ‘s’
* two-phase commit (for reference tables): ‘t’
|
SELECT * from pg_dist_partition;
logicalrelid | partmethod | partkey | colocationid | repmodel
---------------+------------+------------------------------------------------------------------------------------------------------------------------+--------------+----------
github_events | h | {VAR :varno 1 :varattno 4 :vartype 20 :vartypmod -1 :varcollid 0 :varlevelsup 0 :varnoold 1 :varoattno 4 :location -1} | 2 | c
(1 row)
Shard table
The pg_dist_shard table stores metadata about individual shards of a table. This includes information about which distributed table the shard belongs to and statistics about the distribution column for that shard. For append distributed tables, these statistics correspond to min / max values of the distribution column. In case of hash distributed tables, they are hash token ranges assigned to that shard. These statistics are used for pruning away unrelated shards during SELECT queries.
Name |
Type |
Description |
|---|---|---|
logicalrelid |
regclass |
Distributed table to which this shard belongs. This value references the
relfilenode column in the pg_class system catalog table.
|
shardid |
bigint |
Globally unique identifier assigned to this shard.
|
shardstorage |
char |
Type of storage used for this shard. Different storage types are
discussed in the table below.
|
shardminvalue |
text |
For append distributed tables, minimum value of the distribution column
in this shard (inclusive).
For hash distributed tables, minimum hash token value assigned to that
shard (inclusive).
|
shardmaxvalue |
text |
For append distributed tables, maximum value of the distribution column
in this shard (inclusive).
For hash distributed tables, maximum hash token value assigned to that
shard (inclusive).
|
SELECT * from pg_dist_shard;
logicalrelid | shardid | shardstorage | shardminvalue | shardmaxvalue
---------------+---------+--------------+---------------+---------------
github_events | 102026 | t | 268435456 | 402653183
github_events | 102027 | t | 402653184 | 536870911
github_events | 102028 | t | 536870912 | 671088639
github_events | 102029 | t | 671088640 | 805306367
(4 rows)
Shard Storage Types
The shardstorage column in pg_dist_shard indicates the type of storage used for the shard. A brief overview of different shard storage types and their representation is below.
Storage Type |
Shardstorage value |
Description |
|---|---|---|
TABLE |
‘t’ |
Indicates that shard stores data belonging to a regular
distributed table.
|
COLUMNAR |
‘c’ |
Indicates that shard stores columnar data. (Used by
distributed cstore_fdw tables)
|
FOREIGN |
‘f’ |
Indicates that shard stores foreign data. (Used by
distributed file_fdw tables)
|
Shard information view
In addition to the low-level shard metadata table described above, Canopy provides a canopy_shards view to easily check:
Where each shard is (node, and port),
What kind of table it belongs to, and
Its size
This view helps you inspect shards to find, among other things, any size imbalances across nodes.
SELECT * FROM canopy_shards;
.
table_name | shardid | shard_name | canopy_table_type | colocation_id | nodename | nodeport | shard_size
------------+---------+--------------+------------------+---------------+-----------+----------+------------
dist | 102170 | dist_102170 | distributed | 34 | localhost | 9701 | 90677248
dist | 102171 | dist_102171 | distributed | 34 | localhost | 9702 | 90619904
dist | 102172 | dist_102172 | distributed | 34 | localhost | 9701 | 90701824
dist | 102173 | dist_102173 | distributed | 34 | localhost | 9702 | 90693632
ref | 102174 | ref_102174 | reference | 2 | localhost | 9701 | 8192
ref | 102174 | ref_102174 | reference | 2 | localhost | 9702 | 8192
dist2 | 102175 | dist2_102175 | distributed | 34 | localhost | 9701 | 933888
dist2 | 102176 | dist2_102176 | distributed | 34 | localhost | 9702 | 950272
dist2 | 102177 | dist2_102177 | distributed | 34 | localhost | 9701 | 942080
dist2 | 102178 | dist2_102178 | distributed | 34 | localhost | 9702 | 933888
The colocation_id refers to the colocation group. For more info about canopy_table_type, see Table Types.
Shard placement table
The pg_dist_placement table tracks the location of shard replicas on worker nodes. Each replica of a shard assigned to a specific node is called a shard placement. This table stores information about the health and location of each shard placement.
Name |
Type |
Description |
|---|---|---|
placementid |
bigint |
Unique auto-generated identifier for each individual placement.
|
shardid |
bigint |
Shard identifier associated with this placement. This value references
the shardid column in the pg_dist_shard catalog table.
|
shardstate |
int |
Describes the state of this placement. Different shard states are
discussed in the section below.
|
shardlength |
bigint |
For append distributed tables, the size of the shard placement on the
worker node in bytes.
For hash distributed tables, zero.
|
groupid |
int |
Identifier used to denote a group of one primary server and zero or more
secondary servers.
|
SELECT * from pg_dist_placement;
placementid | shardid | shardstate | shardlength | groupid
-------------+---------+------------+-------------+---------
1 | 102008 | 1 | 0 | 1
2 | 102008 | 1 | 0 | 2
3 | 102009 | 1 | 0 | 2
4 | 102009 | 1 | 0 | 3
5 | 102010 | 1 | 0 | 3
6 | 102010 | 1 | 0 | 4
7 | 102011 | 1 | 0 | 4
Shard Placement States
Canopy manages shard health on a per-placement basis and automatically marks a placement as unavailable if leaving the placement in service would put the cluster in an inconsistent state. The shardstate column in the pg_dist_placement table is used to store the state of shard placements. A brief overview of different shard placement states and their representation is below.
State name |
Shardstate value |
Description |
|---|---|---|
FINALIZED |
1 |
This is the state new shards are created in. Shard placements
in this state are considered up-to-date and are used in query
planning and execution.
|
INACTIVE |
3 |
Shard placements in this state are considered inactive due to
being out-of-sync with other replicas of the same shard. This
can occur when an append, modification (INSERT, UPDATE or
DELETE ) or a DDL operation fails for this placement. The query
planner will ignore placements in this state during planning and
execution. Users can synchronize the data in these shards with
a finalized replica as a background activity.
|
TO_DELETE |
4 |
If Canopy attempts to drop a shard placement in response to a
master_apply_delete_command call and fails, the placement is
moved to this state. Users can then delete these shards as a
subsequent background activity.
|
Worker node table
The pg_dist_node table contains information about the worker nodes in the cluster.
Name |
Type |
Description |
|---|---|---|
nodeid |
int |
Auto-generated identifier for an individual node.
|
groupid |
int |
Identifier used to denote a group of one primary server and zero or more
secondary servers. By default it is the same as the nodeid.
|
nodename |
text |
Host Name or IP Address of the LightDB worker node.
|
nodeport |
int |
Port number on which the LightDB worker node is listening.
|
noderack |
text |
(Optional) Rack placement information for the worker node.
|
hasmetadata |
boolean |
Reserved for internal use.
|
isactive |
boolean |
Whether the node is active accepting shard placements.
|
noderole |
text |
Whether the node is a primary or secondary
|
nodecluster |
text |
The name of the cluster containing this node
|
metadatasynced |
boolean |
Reserved for internal use.
|
shouldhaveshards |
boolean |
If false, shards will be moved off node (drained) when rebalancing,
nor will shards from new distributed tables be placed on the node,
unless they are colocated with shards already there
|
SELECT * from pg_dist_node;
nodeid | groupid | nodename | nodeport | noderack | hasmetadata | isactive | noderole | nodecluster | metadatasynced | shouldhaveshards
--------+---------+-----------+----------+----------+-------------+----------+----------+-------------+----------------+------------------
1 | 1 | localhost | 12345 | default | f | t | primary | default | f | t
2 | 2 | localhost | 12346 | default | f | t | primary | default | f | t
3 | 3 | localhost | 12347 | default | f | t | primary | default | f | t
(3 rows)
Distributed object table
The canopy.pg_dist_object table contains a list of objects such as types and functions that have been created on the coordinator node and propagated to worker nodes. When an administrator adds new worker nodes to the cluster, Canopy automatically creates copies of the distributed objects on the new nodes (in the correct order to satisfy object dependencies).
Name |
Type |
Description |
|---|---|---|
classid |
oid |
Class of the distributed object |
objid |
oid |
Object id of the distributed object |
objsubid |
integer |
Object sub id of the distributed object, e.g. attnum |
type |
text |
Part of the stable address used during pg upgrades |
object_names |
text[] |
Part of the stable address used during pg upgrades |
object_args |
text[] |
Part of the stable address used during pg upgrades |
distribution_argument_index |
integer |
Only valid for distributed functions/procedures |
colocationid |
integer |
Only valid for distributed functions/procedures |
“Stable addresses” uniquely identify objects independently of a specific server. Canopy tracks objects during a LightDB upgrade using stable addresses created with the pg_identify_object_as_address() function.
Here’s an example of how create_distributed_function() adds entries to the
canopy.pg_dist_object table:
CREATE TYPE stoplight AS enum ('green', 'yellow', 'red');
CREATE OR REPLACE FUNCTION intersection()
RETURNS stoplight AS $$
DECLARE
color stoplight;
BEGIN
SELECT *
FROM unnest(enum_range(NULL::stoplight)) INTO color
ORDER BY random() LIMIT 1;
RETURN color;
END;
$$ LANGUAGE plpgsql VOLATILE;
SELECT create_distributed_function('intersection()');
-- will have two rows, one for the TYPE and one for the FUNCTION
TABLE canopy.pg_dist_object;
-[ RECORD 1 ]---------------+------
classid | 1247
objid | 16780
objsubid | 0
type |
object_names |
object_args |
distribution_argument_index |
colocationid |
-[ RECORD 2 ]---------------+------
classid | 1255
objid | 16788
objsubid | 0
type |
object_names |
object_args |
distribution_argument_index |
colocationid |
Canopy tables view
The canopy_tables view shows a summary of all tables managed by Canopy (distributed and reference tables). The view combines information from Canopy metadata tables for an easy, human-readable overview of these table properties:
Human-readable size
Shard count
Owner (database user)
Access method (heap or columnar)
Here’s an example:
SELECT * FROM canopy_tables;
┌────────────┬───────────────────┬─────────────────────┬───────────────┬────────────┬─────────────┬─────────────┬───────────────┐
│ table_name │ canopy_table_type │ distribution_column │ colocation_id │ table_size │ shard_count │ table_owner │ access_method │
├────────────┼───────────────────┼─────────────────────┼───────────────┼────────────┼─────────────┼─────────────┼───────────────┤
│ foo.test │ distributed │ test_column │ 1 │ 0 bytes │ 32 │ canopy │ heap │
│ ref │ reference │ <none> │ 2 │ 24 GB │ 1 │ canopy │ heap │
│ test │ distributed │ id │ 1 │ 248 TB │ 32 │ canopy │ heap │
└────────────┴───────────────────┴─────────────────────┴───────────────┴────────────┴─────────────┴─────────────┴───────────────┘
Time partitions view
Canopy provides UDFs to manage partitions for the Timeseries Data use case.
It also maintains a time_partitions view to inspect the partitions it
manages.
Columns:
parent_table the table which is partitioned
partition_column the column on which the parent table is partitioned
partition the name of a partition table
from_value lower bound in time for rows in this partition
to_value upper bound in time for rows in this partition
access_method
heapfor row-based storage, andcolumnarfor columnar storage
SELECT * FROM time_partitions;
┌────────────────────────┬──────────────────┬─────────────────────────────────────────┬─────────────────────┬─────────────────────┬───────────────┐
│ parent_table │ partition_column │ partition │ from_value │ to_value │ access_method │
├────────────────────────┼──────────────────┼─────────────────────────────────────────┼─────────────────────┼─────────────────────┼───────────────┤
│ github_columnar_events │ created_at │ github_columnar_events_p2015_01_01_0000 │ 2015-01-01 00:00:00 │ 2015-01-01 02:00:00 │ columnar │
│ github_columnar_events │ created_at │ github_columnar_events_p2015_01_01_0200 │ 2015-01-01 02:00:00 │ 2015-01-01 04:00:00 │ columnar │
│ github_columnar_events │ created_at │ github_columnar_events_p2015_01_01_0400 │ 2015-01-01 04:00:00 │ 2015-01-01 06:00:00 │ columnar │
│ github_columnar_events │ created_at │ github_columnar_events_p2015_01_01_0600 │ 2015-01-01 06:00:00 │ 2015-01-01 08:00:00 │ heap │
└────────────────────────┴──────────────────┴─────────────────────────────────────────┴─────────────────────┴─────────────────────┴───────────────┘
Co-location group table
The pg_dist_colocation table contains information about which tables’ shards should be placed together, or co-located. When two tables are in the same co-location group, Canopy ensures shards with the same partition values will be placed on the same worker nodes. This enables join optimizations, certain distributed rollups, and foreign key support. Shard co-location is inferred when the shard counts, replication factors, and partition column types all match between two tables; however, a custom co-location group may be specified when creating a distributed table, if so desired.
Name |
Type |
Description |
|---|---|---|
colocationid |
int |
Unique identifier for the co-location group this row corresponds to.
|
shardcount |
int |
Shard count for all tables in this co-location group
|
replicationfactor |
int |
Replication factor for all tables in this co-location group.
|
distributioncolumntype |
oid |
The type of the distribution column for all tables in this
co-location group.
|
distributioncolumncollation |
oid |
The collation of the distribution column for all tables in
this co-location group.
|
SELECT * from pg_dist_colocation;
colocationid | shardcount | replicationfactor | distributioncolumntype | distributioncolumncollation
--------------+------------+-------------------+------------------------+-----------------------------
2 | 32 | 2 | 20 | 0
(1 row)
Rebalancer strategy table
This table defines strategies that rebalance_table_shards can use to determine where to move shards.
Name |
Type |
Description |
|---|---|---|
name |
name |
Unique name for the strategy
|
default_strategy |
boolean |
Whether rebalance_table_shards should choose this strategy by
default. Use canopy_set_default_rebalance_strategy to update
this column
|
shard_cost_function |
regproc |
Identifier for a cost function, which must take a shardid as bigint,
and return its notion of a cost, as type real
|
node_capacity_function |
regproc |
Identifier for a capacity function, which must take a nodeid as int,
and return its notion of node capacity as type real
|
shard_allowed_on_node_function |
regproc |
Identifier for a function that given shardid bigint, and nodeidarg int,
returns boolean for whether the shard is allowed to be stored on the
node
|
default_threshold |
float4 |
Threshold for deeming a node too full or too empty, which determines
when the rebalance_table_shards should try to move shards
|
minimum_threshold |
float4 |
A safeguard to prevent the threshold argument of
rebalance_table_shards() from being set too low
|
improvement_threshold |
float4 |
Determines when moving a shard is worth it during a rebalance.
The rebalancer will move a shard when the ratio of the improvement with
the shard move to the improvement without crosses the threshold. This
is most useful with the by_disk_size strategy.
|
A Canopy installation ships with these strategies in the table:
SELECT * FROM pg_dist_rebalance_strategy;
-[ RECORD 1 ]------------------+---------------------------------
name | by_shard_count
default_strategy | t
shard_cost_function | canopy_shard_cost_1
node_capacity_function | canopy_node_capacity_1
shard_allowed_on_node_function | canopy_shard_allowed_on_node_true
default_threshold | 0
minimum_threshold | 0
improvement_threshold | 0
-[ RECORD 2 ]------------------+---------------------------------
name | by_disk_size
default_strategy | f
shard_cost_function | canopy_shard_cost_by_disk_size
node_capacity_function | canopy_node_capacity_1
shard_allowed_on_node_function | canopy_shard_allowed_on_node_true
default_threshold | 0.1
minimum_threshold | 0.01
improvement_threshold | 0.5
The default strategy, by_shard_count, assigns every shard the same cost. Its effect is to equalize the shard count across nodes. The other predefined strategy, by_disk_size, assigns a cost to each shard matching its disk size in bytes plus that of the shards that are colocated with it. The disk size is calculated using pg_total_relation_size, so it includes indices. This strategy attempts to achieve the same disk space on every node. Note the threshold of 0.1 – it prevents unnecessary shard movement caused by insigificant differences in disk space.
Distributed Query Activity
In some situations, queries might get blocked on row-level locks on one of the shards on a worker node. If that happens then those queries would not show up in pg_locks on the Canopy coordinator node.
Canopy provides special views to watch queries and locks throughout the cluster, including shard-specific queries used internally to build results for distributed queries.
canopy_dist_stat_activity: shows the distributed queries that are executing on all nodes. A superset of
pg_stat_activity, usable wherever the latter is.canopy_worker_stat_activity: shows queries on workers, including fragment queries against individual shards.
canopy_lock_waits: Blocked queries throughout the cluster.
The first two views include all columns of pg_stat_activity plus the host/port of the worker that initiated the query and the host/port of the coordinator node of the cluster.
For example, consider counting the rows in a distributed table:
-- run from worker on localhost:9701
SELECT count(*) FROM users_table;
We can see the query appear in canopy_dist_stat_activity:
SELECT * FROM canopy_dist_stat_activity;
-[ RECORD 1 ]----------+----------------------------------
query_hostname | localhost
query_hostport | 9701
master_query_host_name | localhost
master_query_host_port | 9701
transaction_number | 1
transaction_stamp | 2018-10-05 13:27:20.691907+03
datid | 12630
datname | postgres
pid | 23723
usesysid | 10
usename | canopy
application_name | ltsql
client_addr |
client_hostname |
client_port | -1
backend_start | 2018-10-05 13:27:14.419905+03
xact_start | 2018-10-05 13:27:16.362887+03
query_start | 2018-10-05 13:27:20.682452+03
state_change | 2018-10-05 13:27:20.896546+03
wait_event_type | Client
wait_event | ClientRead
state | idle in transaction
backend_xid |
backend_xmin |
query | SELECT count(*) FROM users_table;
backend_type | client backend
This query requires information from all shards. Some of the information is in shard users_table_102038 which happens to be stored in localhost:9700. We can see a query accessing the shard by looking at the canopy_worker_stat_activity view:
SELECT * FROM canopy_worker_stat_activity;
-[ RECORD 1 ]----------+-----------------------------------------------------------------------------------------
query_hostname | localhost
query_hostport | 9700
master_query_host_name | localhost
master_query_host_port | 9701
transaction_number | 1
transaction_stamp | 2018-10-05 13:27:20.691907+03
datid | 12630
datname | postgres
pid | 23781
usesysid | 10
usename | canopy
application_name | canopy
client_addr | ::1
client_hostname |
client_port | 51773
backend_start | 2018-10-05 13:27:20.75839+03
xact_start | 2018-10-05 13:27:20.84112+03
query_start | 2018-10-05 13:27:20.867446+03
state_change | 2018-10-05 13:27:20.869889+03
wait_event_type | Client
wait_event | ClientRead
state | idle in transaction
backend_xid |
backend_xmin |
query | COPY (SELECT count(*) AS count FROM users_table_102038 users_table WHERE true) TO STDOUT
backend_type | client backend
The query field shows data being copied out of the shard to be counted.
Note
If a router query (e.g. single-tenant in a multi-tenant application, SELECT * FROM table WHERE tenant_id = X) is executed without a transaction block, then canopy_query_host_name and canopy_query_host_port columns will be NULL in canopy_worker_stat_activity.
Here are examples of useful queries you can build using
canopy_worker_stat_activity:
-- active queries' wait events on a certain node
SELECT query, wait_event_type, wait_event
FROM canopy_worker_stat_activity
WHERE query_hostname = 'xxxx' and state='active';
-- active queries' top wait events
SELECT wait_event, wait_event_type, count(*)
FROM canopy_worker_stat_activity
WHERE state='active'
GROUP BY wait_event, wait_event_type
ORDER BY count(*) desc;
-- total internal connections generated per node by Canopy
SELECT query_hostname, count(*)
FROM canopy_worker_stat_activity
GROUP BY query_hostname;
-- total internal active connections generated per node by Canopy
SELECT query_hostname, count(*)
FROM canopy_worker_stat_activity
WHERE state='active'
GROUP BY query_hostname;
The next view is canopy_lock_waits. To see how it works, we can generate a locking situation manually. First we’ll set up a test table from the coordinator:
CREATE TABLE numbers AS
SELECT i, 0 AS j FROM generate_series(1,10) AS i;
SELECT create_distributed_table('numbers', 'i');
Then, using two sessions on the coordinator, we can run this sequence of statements:
-- session 1 -- session 2
------------------------------------- -------------------------------------
BEGIN;
UPDATE numbers SET j = 2 WHERE i = 1;
BEGIN;
UPDATE numbers SET j = 3 WHERE i = 1;
-- (this blocks)
The canopy_lock_waits view shows the situation.
SELECT * FROM canopy_lock_waits;
-[ RECORD 1 ]-------------------------+----------------------------------------
waiting_pid | 88624
blocking_pid | 88615
blocked_statement | UPDATE numbers SET j = 3 WHERE i = 1;
current_statement_in_blocking_process | UPDATE numbers SET j = 2 WHERE i = 1;
waiting_node_id | 0
blocking_node_id | 0
waiting_node_name | coordinator_host
blocking_node_name | coordinator_host
waiting_node_port | 5432
blocking_node_port | 5432
In this example the queries originated on the coordinator, but the view can also list locks between queries originating on workers.