From: Mason S. <ms...@tr...> - 2014-07-02 00:00:32
|
On Tue, Jul 1, 2014 at 6:59 AM, Wilkerson, Daniel <dwi...@fu...> wrote: > Hey everyone. I have my cluster up and running using pgxc_ctl instead of > manually configuring (XC 1.1; Ubuntu 13.04). It's definitely easier to > setup and control the cluster this way for sure. I have HAProxy balancing > the requests to the servers. It's working fine also. > > All of the nodes startup fine and everything seems ok on the surface; that > is until I run the DDL script to create my database across all the nodes. > During script execution I get, "ERROR: sorry, too many clients already". > I've tried upping the number of available connections but still get the > same issue across the cluster. I am executing the DDL against my first > coordinator. > > Could this be a mismatch in my config of the proportions between shared > mem and max connections? I've queried the database and it doesn't report > tons of connections. I'm not sure where to look in my config to suss out > the problem. > > I appreciate any light you all could shed on this. I'm fairly certain it's > something I need to tweak in my config somewhere, but also want to make > sure I've not run up against a bug. > > Thanks! > dw > > > Connection values entered into the pgxc_ctl cluster config: > max_connections = 200 > max_pool_size = 200 > These should be appended in the individual nodes' postgresql.conf files. I would double-check there, adding them manually if necessary. > > This is a snippet from the script erring out. > … > CREATE TABLE > CREATE SEQUENCE > ALTER SEQUENCE > CREATE TABLE > CREATE SEQUENCE > ALTER SEQUENCE > CREATE TABLE > CREATE SEQUENCE > ALTER SEQUENCE > psql:db.sql:2654: WARNING: unexpected EOF on datanode connection > psql:db.sql:2654: WARNING: unexpected EOF on datanode connection > psql:db.sql:2654: ERROR: sorry, too many clients already > > > This snippet from ps –ef | grep post on my first coordinator/datanode > combo shows a hung create table statement: > > postgre+ 3608 3588 0 09:39 pts/1 00:00:00 psql -p 20004 -d postgres > postgre+ 3616 2980 0 09:40 ? 00:00:00 postgres: postgres-xc db > [local] CREATE TABLE > postgre+ 3644 3291 0 09:41 ? 00:00:00 postgres: checkpointer > process > postgre+ 3645 3291 0 09:41 ? 00:00:00 postgres: writer process > postgre+ 3646 3291 0 09:41 ? 00:00:00 postgres: wal writer > process > postgre+ 3647 3291 0 09:41 ? 00:00:00 postgres: autovacuum > launcher process > postgre+ 3648 3291 0 09:41 ? 00:00:00 postgres: stats collector > process > postgre+ 3705 2553 0 09:48 pts/0 00:00:00 ps -ef > postgre+ 3706 2553 0 09:48 pts/0 00:00:00 grep post > OTOH, this does not look like very much activity at all... there should be plenty of connections. -- Mason Sharp TransLattice - http://www.translattice.com Distributed and Clustered Database Solutions |