[PERFORM] Postgres installation for Performance

2009-06-06 Thread S Arvind
In single server(4gb ram 2 core 2 duo), which design is good for performance? 1) single installation for 200 db with 800 tables/db or 2) Two installation (instance) running on different port with each handling 100 db. Which of this design is good for postgres where our goal is high perfor

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread S Arvind
Thanks Tom Lane, I think we must have to consider about your last mail words. But now reducing the table is mearly impossible, but very thanks for advice , we will try it in future. -Arvind S On Sun, Jun 7, 2009 at 4:42 AM, Tom Lane wrote: > S Arvind writes: > > So do i have to increase t

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread Tom Lane
S Arvind writes: > So do i have to increase the max_fsm_relation based on (Average_no_relation > per db * number of db)? if so it will be very high since in our one db > server we have 200 db with average 800 tables in each db. What is the value > we have to give for this kind of server? About 16

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread S Arvind
Sorry Tom, i cant able to understand. Should i have to increse the max_fsm_rel based on formula and re-run the vacuum command? The main reason for vacuum for us is to increase performance of our db. Please tell value for our kind of server(as provided in previous mail) ? -- Arvind S On Sun, Jun 7

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread Tom Lane
S Arvind writes: > So is it no use running > vacuumdb --all --analyze --full > as fsm map is full? Well, it's not of *no* use. But you'd be well advised to crank up the FSM size. regards, tom lane -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread S Arvind
Thanks Tom, So do i have to increase the max_fsm_relation based on (Average_no_relation per db * number of db)? if so it will be very high since in our one db server we have 200 db with average 800 tables in each db. What is the value we have to give for this kind of server? -Arvind S On Sun, J

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread S Arvind
So is it no use running vacuumdb --all --analyze --full as fsm map is full? -Arvind S On Sun, Jun 7, 2009 at 4:24 AM, S Arvind wrote: > Thanks Tom, > So do i have to increase the max_fsm_relation based on (Average_no_relation > per db * number of db)? if so it will be very high since in our on

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread Tom Lane
S Arvind writes: > But there nearly only 300 tables in that db. Is the free space map is per > DB or for all DB. Can i know the reason of this problem? It's across all DBs in the installation. regards, tom lane -- Sent via pgsql-performance mailing list (pgsql-performa

Re: [PERFORM] Vacuum ALL FULL

2009-06-06 Thread S Arvind
should i have to increase max_fsm_relations more. If yes why i have to ? Since number of relation is less only. --Arvind S On Sun, Jun 7, 2009 at 4:06 AM, S Arvind wrote: > Found a notice after completing the > vacuumdb -p 5433 -- all --analyze --full -v > max_fsm_relation = 1400 in postgre

[PERFORM] Vacuum ALL FULL

2009-06-06 Thread S Arvind
Found a notice after completing the vacuumdb -p 5433 -- all --analyze --full -v max_fsm_relation = 1400 in postgresql.conf Thou all our 50 db individually have less then 1400 relation , when it completes , there was NOTICE that increase the max_fsm_relation. INFO: free space map contains 10344 p

Re: [PERFORM] Pointers needed on optimizing slow SQL statements

2009-06-06 Thread Simon Riggs
On Wed, 2009-06-03 at 21:21 -0400, Robert Haas wrote: > But, we're not always real clever about selectivity. Sometimes you > have to fake the planner out, as discussed here. > > http://archives.postgresql.org/pgsql-performance/2009-06/msg00023.php > > Actually, I had to do this today on a prod