Subject Re: AW: AW: [firebird-support] Problem Firebird Classic Server
Author Helen Borrie
At 07:29 PM 11/06/2008, you wrote:
>Hello Anderson, Helen and Fabian,
>now I'm confused.
>First my gstat -h statistics:
>Flags: 0
>Checksum 12345
>Generation 230320
>Page Size 4096
>ODS version 11.0
>Oldest transaction 833
>Oldest active 961
>Oldest snapshot 840
>Next transaction 229603
>Bumped transaction 1
>Sequence number 0
>Next attached ID 711
>Implementation ID 16
>Shadow count 0
>Page buffers 0

- this means that the value used in firebird.conf will be used for *each connection*. You have 2048. Too much.

>Next header page 0
>Database dialect 3
>Creation Date Jun 6, 2008 10:31:39
>Sweep interval: 20000
>The server has 4 GB of RAM.
>Firebird.conf: #DefaultDbCachePages = 2048 - I don't have changed at the

It must be changed. Delete the # symbol and change the number to 128.

>I have read all your responses.
>It is right, that I change the firebird.conf - value #DefaultDbCachePages =
>2048 to DefaultDbCachePages = 75 ? or better 128?

128 is a reasonable size, considering your page size of 4 Kb and that you have 4 GB of RAM. Try it. Try 75 and 256 too, if you like, and compare.

Don't forget that you must get all users to close their connections before the change will start to take effect. If you are using Classic on Windows you will need to stop and restart the fb_inet_server service as well.

> I have three suggestions and different values.

All are suggestions. Try all three and observe which one gives you the best improvement.

>Shall I change also the Page Size?

No. There is no reason to change the page size.

> Are the many error-messages a result of the wrong values? (FLS (Client) Thu
>Dec 06 08:41:50 2007
> INET/inet_error: connect errno = 10060)

Indirectly, yes. When a process cannot get enough RAM to execute a request, it will crash. Then, when the client makes another request, there is no server process at the end of the wire and the client gets a 10060 error from the network.

> I am unsure about at the moment because there are different opinions.

You are receiving different pieces of information relating to your problem and all are relevant. OK, Fabian Chocron's theory about memory usage was not correct....but everyone is telling you that Superserver cache settings are wrong for Classic. It is documented in a lot of places....