|
1 | 1 | # `multimaster` |
2 | 2 |
|
3 | | -A synchronous multi-master replication based on **snapshot sharing**. |
| 3 | +An implementation of synchronous **multi-master replication** based on **snapshot sharing**. |
4 | 4 |
|
5 | | -## Installing |
| 5 | +## Usage |
6 | 6 |
|
7 | | -1. Build and install postgres from this repo on all machines in cluster. |
8 | | -1. Install contrib/raftable and contrib/mmts extensions. |
9 | | -1. Right now we need clean postgres installation to spin up multimaster cluster. |
10 | | -1. Create required database inside postgres before enabling multimaster extension. |
11 | | -1. We are requiring following postgres configuration: |
12 | | - * 'max_prepared_transactions' > 0 -- in multimaster all writing transaction along with ddl are wrapped as two-phase transaction, so this number will limit maximum number of writing transactions in this cluster node. |
13 | | - * 'synchronous_commit - off' -- right now we do not support async commit. (one can enable it, but that will not bring desired effect) |
14 | | - * 'wal_level = logical' -- multimaster built on top of logical replication so this is mandatory. |
15 | | - * 'max_wal_senders' -- this should be at least number of nodes - 1 |
16 | | - * 'max_replication_slots' -- this should be at least number of nodes - 1 |
17 | | - * 'max_worker_processes' -- at least 2*N + 1 + P, where N is number of nodes in cluster, P size of pool of workers(see below) (1 raftable, n-1 receiver, n-1 sender, mtm-sender, mtm-receiver, + number of pool worker). |
18 | | - * 'default_transaction_isolation = 'repeatable read'' -- multimaster isn't supporting default read commited level. |
19 | | -1. Multimaster have following configuration parameters: |
20 | | - * 'multimaster.conn_strings' -- connstrings for all nodes in cluster, separated by comma. |
21 | | - * 'multimaster.node_id' -- id of current node, number starting from one. |
22 | | - * 'multimaster.workers' -- number of workers that can apply transactions from neighbouring nodes. |
23 | | - * 'multimaster.use_raftable = true' -- just set this to true. Deprecated. |
24 | | - * 'multimaster.queue_size = 52857600' -- queue size for applying transactions from neighbouring nodes. |
25 | | - * 'multimaster.ignore_tables_without_pk = 1' -- do not replicate tables without primary key |
26 | | - * 'multimaster.heartbeat_send_timeout = 250' -- heartbeat period (ms). |
27 | | - * 'multimaster.heartbeat_recv_timeout = 1000' -- disconnect node if we miss heartbeats all that time (ms). |
28 | | - * 'multimaster.twopc_min_timeout = 40000' -- rollback stalled transaction after this period (ms). |
29 | | - * 'raftable.id' -- id of current node, number starting from one. |
30 | | - * 'raftable.peers' -- id of current node, number starting from one. |
31 | | -1. Allow replication in pg_hba.conf. |
| 7 | +1. Install `contrib/arbiter` and `contrib/multimaster` on each instance. |
| 8 | +1. Add these required options to the `postgresql.conf` of each instance in the cluster. |
32 | 9 |
|
33 | | -## Status functions |
| 10 | + ```sh |
| 11 | + multimaster.workers = 8 |
| 12 | + multimaster.queue_size = 10485760 # 10mb |
| 13 | + multimaster.local_xid_reserve = 100 # number of xids reserved for local transactions |
| 14 | + multimaster.buffer_size = 0 # sockhub buffer size, if 0, then direct connection will be used |
| 15 | + multimaster.arbiters = '127.0.0.1:5431,127.0.0.1:5430' |
| 16 | + # comma-separated host:port pairs where arbiters reside |
| 17 | + multimaster.conn_strings = 'replication=database dbname=postgres ...' |
| 18 | + # comma-separated list of connection strings |
| 19 | + multimaster.node_id = 1 # the 1-based index of the node in the cluster |
| 20 | + shared_preload_libraries = 'multimaster' |
| 21 | + max_connections = 200 |
| 22 | + max_replication_slots = 10 # at least the number of nodes |
| 23 | + wal_level = logical # multimaster is build on top of |
| 24 | + # logical replication and will not work otherwise |
| 25 | + max_worker_processes = 100 # at least (FIXME: need an estimation here) |
| 26 | + ``` |
34 | 27 |
|
35 | | -* mtm.get_nodes_state() -- show status of nodes on cluster |
36 | | -* mtm.get_cluster_state() -- show whole cluster status |
37 | | -* mtm.get_cluster_info() -- print some debug info |
38 | | -* mtm.make_table_local(relation regclass) -- stop replication for a given table |
| 28 | +## Testing |
| 29 | + |
| 30 | +1. `cd contrib/multimaster` |
| 31 | +1. Deploy the cluster somewhere. You can use `tests/daemons.go` or one of `tests/*.sh` for that. |
| 32 | +1. `make -C tests` |
| 33 | +1. `tests/dtmbench ...`. See `tests/run.sh` for an example. |
0 commit comments