--- In firebird-support@yahoogroups.com, "Kevin Stanton"
<Kevin.Stanton@...> wrote:

>
> Good Day,
> I have recently (within the last couple of months) installed

Firebird 1.5 on

> a duel processor system running FB Classic. The server is running MS
> Windows 2003 server and has a ton of disk space and 3 GB of RAM. I

believe

> it is HP hardware running Raid 5 with duel processors.
>
> I've gotten a few reports since install that my application just

"freezes".

> Some reports can be run but data entry freezes up (order entry, check
> writing, etc). The database itself is getting large (to me anyway)

at about

> 1.8 GB, dialect 1, forced writes are on as well as OldParamOrdering

is on

> due to my still using an older version of IBO.
>
> I have not been able to dial in to check the FB log yet.
>
> I did dial in this AM, ran a "mend" batch file which is an excerpt from
> Helen's book to validate, mend, validate, backup and restore the

database

> from a copy of the live one. This seems to fix the freezing problem.

Until there is more information available, it is all guessing, but
Superserver has a garbage collection thread, Classic does not. With
classic, it is the poor guy that is in the wrong place at the wrong
time who gets handed the mop and told to clean up. Although IIRC this
still happens with SS, there is a thread dedicated to the task of
background garbage collection so I would expect that the garbage would
have less chance to accumulate under SS.

A backup-restore cycle will obliterate any garbage lingering around.
Do you do a daily backup with gbak or services manager (because this
as a bonus does gc as it runs)?

Are there any actions of your application that are particularly
garbage prone? Firebird attempts to undo all actions when you rollback
nicely, but if you abandon the transaction (killing the process for
example), all the uncommitted work will have to be cleaned up by
someone else.

Problems with garbage tend to occur where you have indices involved
with a lot of duplicate values. This particular case is largely fixed
in FB 2, but is worth knowing about in Fb 1.5. You can (sort of but
not really safely) use triggers to enforce a FK relationship with poor
selectivity (as the automatically generated index of a FK relationship
often has a lot of duplicates).

>
> Has anyone run into something like this before? This client used to

be on a

> single processor machine running FB SS and we never ran into this

situation.

> According to their hardware support folk, there are no reports of disk
> errors.
>
> I'm contemplating uninstalling FB CS and installing FB SS with the

affinity

> set to 1 CPU to see if this is the problem.

I switched server last night (the other way though) in under 2
minutes, so if you have enough possible downtime to run your
mend-validate-backup-restore script you mention, you certainly have
enough time to switch server modes.

Adam