与 Azure Postgres 的连接时间长

Question

我有 Azure Database for PostgreSQL 服务 (PaaS)。当我尝试使用psql 查询它时，即使是来自一个表的简单SELECT 查询也需要大约1.5 秒。当我在 postgres 控制台中时，没有问题，查询在 100 毫秒内执行。什么会导致连接时间延迟？

for I in {1..15} ; do /usr/bin/time -f "%es" psql -U postgres@user  -h  my-server.postgres.database.azure.com database_name -c "SELECT count(*) FROM table_name" > /dev/null ; done
1.86s
1.95s
1.60s
1.03s
1.59s
1.58s
1.60s
1.68s
1.60s
1.60s
1.62s
1.63s
1.61s
1.83s
1.70s

仅供参考：表中有 11 个条目

更新 下面的测试证明这是连接延迟而不是在 postgres 上执行查询本身

$ psql -t -U postgres@resource-group  -h  postgresql-project.postgres.database.azure.com stage_table <<EOF
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
SELECT count(*), clock_timestamp() FROM my_table;
EOF
    17 | 2019-07-23 13:26:11.475655+00
    17 | 2019-07-23 13:26:11.506887+00
    17 | 2019-07-23 13:26:11.553769+00
    17 | 2019-07-23 13:26:11.585022+00
    17 | 2019-07-23 13:26:11.616273+00
    17 | 2019-07-23 13:26:11.64752+00
    17 | 2019-07-23 13:26:11.678787+00
    17 | 2019-07-23 13:26:11.710046+00
    17 | 2019-07-23 13:26:11.74129+00
    17 | 2019-07-23 13:26:11.772524+00
    17 | 2019-07-23 13:26:11.803794+00
    17 | 2019-07-23 13:26:11.835042+00
    17 | 2019-07-23 13:26:11.86629+00
    17 | 2019-07-23 13:26:11.897517+00
    17 | 2019-07-23 13:26:11.944397+00
$ 

我还在 azure 上阅读了有关 accelerated networking 的信息，但这看起来仅适用于 VM 本身。我可以在 PostgreSQL 服务上启用此加速吗？ https://docs.microsoft.com/en-us/azure/virtual-network/create-vm-accelerated-networking-cli

Answer 1

延迟是建立数据库连接所需的时间。

1.5 秒相当蹩脚。但它不一定是杀手，因为您应该使用连接池并保持数据库会话打开。

可以肯定的是，你可以将结果与

psql -U postgres@user -h  my-server.postgres.database.azure.com database_name <<EOF
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
SELECT count(*), clock_timestamp() FROM table_name;
EOF

那应该更快。区别在于连接时间。