Re: TRUNCATE HANGS

tmoore  wrote:

> I'm a little tired. I just indicated autocommit false, that was
> incorrect.  Autocommit is true.

I would double-check that.  As I said, what you showed us indicates
that something logging in through TCP on localhost is using
transactions, and not committing them when needed.

> Three threads of executions, java writer, java reader, and psql via
> cron for truncation. As an experiment, the java reader was disabled
> and the test did not block.

TRUNCATE will block or be blocked by just about anything, read or
write.  A reading transaction won't release the conflicting lock
until the transaction ends.

-Kevin