JDBC executeBatch() 在 PostgreSQL 中没有错误地挂起

Question

I try load 50000 rows in table with 200 columns. I do executeBatch() every 1000 rows. And I get lock for this table. The same code works for MS SQL and Oracle but with postgresql I get this issue. When I decrease executeBatch number from 1000 to 75 all works correctly.

Is there any param in config file witch responding for batch buffer size?

Same issue http://www.postgresql.org/message-id/c44a1bc0-dcb1-4b57-8106-e50f9303b7d1@79g2000hsk.googlegroups.com

Answer 1

When I execute insert statements in batch for tables with a large number of columns occurs hanging, when call statement.executeBatch(). It is specific for postgresql jdbc driver. To avoid this issue we should increase buffer size params(SO_SNDBUF, SO_RCVBUF) for socket.

For Windows we have to create such params in register:

[HKEY_LOCAL_MACHINE \SYSTEM \CurrentControlSet \Services \Afd \Parameters]

DefaultReceiveWindow(type=DWORD) = 1640960(Decimal) DefaultSendWindow(type=DWORD) = 1640960(Decimal)

This number(1640960) I get from internet as general recommendation! And it works for me.

Answer 2

Generally you need to look for the following things.

1. Are you able to actually get a lock for the table?
2. Do you have other java code locks that you are waiting for?

In general the first place to check is the pg_stat_activity system view which will show you the query being executed and whether it is active, idle, waiting, etc. Then if it is waiting (ie waiting is t ), then you will want to check the pg_locks view and see what else may have a lock on anything in the relation.

If waiting is not true, then you are better off checking your java code for client-side locks but since this works for MySQL and Oracle, I assume this is less of an issue.

There is one other thing to be aware of. I am not sure about MySQL and Oracle here but PostgreSQL limits you to one query at a time per connection. You might have some locking there?