Thread: One store, two machines?
Greetings. We would like to implement a PostgreSQL database in such a way that the data are housed on disk volumes which are redundant across two Solaris (or maybe Linux) hosts. Both hosts must have write access to the store -- a master system/backup system scheme is not acceptable for us; was also want this to be a load balancing setup. Does Postgres have a facility for keeping data stores on two hosts in sync? TIA, -m -- Michael Jinks, IB Systems Administrator, CCCP finger mjinks@embley.spc.uchicago.edu for public key A guiding voice for all humanity: http://www.yellow5.com/pokey/
Hi out there. Since this morning I'm getting this error when I try to connect to our DB-Server: Sorry, too many clients already What does that mean ? Where can I set this limit ? What's the cause for this error ? I didn't have this before... Greetings /FMK
On Wed, 26 Jul 2000, Florian Meyer-Kassel wrote: > Hi out there. > > Since this morning I'm getting this error when I try to connect to our > DB-Server: > > Sorry, too many clients already > > > What does that mean ? Where can I set this limit ? What's the cause for this > error ? I didn't have this before... > PostgreSQL as normally distributed only allows 32 concurrent users. You have exceeded this number of users. You can fix this in one of two ways. The hard way is to reconfigure & recompile PostgreSQL using the --with-maxbackends=1024 (maximum number of users is 1024!). The simple way is to edit the PostgreSQL startup script to include the parameter "-N 1024" to set the maximum number of concurrent connections to 1024. I don't see any command to list who is currently using PostgreSQL. Hope this helps some, John
Yeah, its called RAID ;) seriously, consider using a Mylex external controller that has twin SCSI bus interfaces on it that allows for multiple hosts the other option would be to write a trigger mechanism implemented in the database to do the writes On Wed, 26 Jul 2000, Michael R. Jinks wrote: > Greetings. > > We would like to implement a PostgreSQL database in such a way that the data > are housed on disk volumes which are redundant across two Solaris (or maybe > Linux) hosts. Both hosts must have write access to the store -- a master > system/backup system scheme is not acceptable for us; was also want this to > be a load balancing setup. > > Does Postgres have a facility for keeping data stores on two hosts in sync? > > TIA, > -m > -- > Michael Jinks, IB > Systems Administrator, CCCP > finger mjinks@embley.spc.uchicago.edu for public key > A guiding voice for all humanity: http://www.yellow5.com/pokey/ > ----------------------------------------------------------------------------- david@backpack.com BackPack Software, Inc. www.backpack.com +1 651.645.7550 voice "Life is an Adventure. +1 651.645.9798 fax Don't forget your BackPack!" -----------------------------------------------------------------------------
John McKown <jmckown@prodigy.net> writes:
> PostgreSQL as normally distributed only allows 32 concurrent users. You
> have exceeded this number of users. You can fix this in one of two ways.
> The hard way is to reconfigure & recompile PostgreSQL using the
> --with-maxbackends=1024 (maximum number of users is 1024!). The simple way
> is to edit the PostgreSQL startup script to include the parameter
> "-N 1024" to set the maximum number of concurrent connections to 1024.
Note you will also need to increase -B (to at least twice -N).
It may be a bad idea to raise the limit as far as 1024, unless your
platform is really configured to support that many active backends
(which means LOTS of RAM and swap space, lots of kernel open-file-table
entries, etc). The reason the default is 32 is that that roughly
corresponds to the factory-stock limits in many Unixen about size of
shared memory blocks and so forth. Don't be surprised if you have to
rejigger kernel parameter settings before you can go up very far.
> I don't see any command to list who is currently using PostgreSQL.
"ps -aux | grep postgres" (adapt to local situation...)
regards, tom lane