Subject | Re: Firebird Server Memory is Increasing |
---|---|
Author | Chooi-Ting |
Post date | 2003-07-28T07:29:10Z |
--- In firebird-support@yahoogroups.com, Helen Borrie <helebor@t...>
wrote:
I am worry because some of my client complain that the firebird
server is hanging and the update/insert process that a longer time.
I am afraid if the memory keep increasing, what will happen to the
firebird server.
monitor the CPU usage, RAM, SWAP of the ibserver process.
wrote:
> At 05:00 AM 17/07/2003 +0000, you wrote:<helebor@t...>
> >--- In firebird-support@yahoogroups.com, Helen Borrie
> >wrote:never
> > > Chooi-Ting,
> > > At 11:18 AM 16/07/2003 +0000, you wrote:
> > > >How to resolve firebird server memory ? it is increasing and
> > > >drop, though the increase number of not much, eg. from 23040to
> > > >23496 in few hours time.needs "resolving"?
> > >
> > > Please explain what you think is a problem. What
I am worry because some of my client complain that the firebird
server is hanging and the update/insert process that a longer time.
I am afraid if the memory keep increasing, what will happen to the
firebird server.
> >ps on
> >Server Version : firebird 1.0
> >architecture: how to check ?
>
> Did you install Superserver or Classic? I'm not sure if you have
> Solaris. If not, use the equivalent command to show you whatprocesses are
> running. If you have a number of processes called gds_inet_server(one per
> attached client) then you have Classic; if you have processescalled
> ibserver, then it is Superserver.it is superserver.
>it
> >platform : Sun Solaris 2.8
> ># of client attached: <10
> >Action: insert/update/delete
> >size of database cache : is it same as allocated db pages? if yes,
> >is 37980waaaaaay
>
> Where did you get this figure of 37980? If you configured
> database_cache_pages (in isc_config) to be this high then it is
> too high, even for SS. Classic would allocate about 300 Mb ofstatic RAM
> to each client connection. SS would make a 300 Mb pooled cache forall
> connections and grow it dynamically if more is needed. It's stilltoo big.
>what
> I think the number you give here probably isn't the right one. See
> you have in isc_config and also check it in isql using SET STATS.The
> defaults are 75 for Classic and 2048 for SS. If this parameter isyour cache
> commented out (#) then you are running with the default.
>
> ISQL> CONNECT database_name;
> ISQL> SET STATS ON;
> ISQL> COMMIT;
> Current memory = 415768
> Delta memory = 2048
> Max memory = 419840
> Elapsed time = 0.03 sec
> Buffers = 2048 <---- this one tells you how many database pages
> is. Cache = this figure times the page_size (in bytes)the buffer is 2048
> Reads = 0
> Writes 2
> Fetches = 2
>
>
> >database page_size : 8192figures a
> >
> >RAM reading now is 23928 (the reading that i captured yesterday is
> >23496).
>
> Where are you reading it from? What is it? and why are these
> problem?the reading is from sun solaris process manager, where you can
monitor the CPU usage, RAM, SWAP of the ibserver process.
>day
>
> >it is a problem if the memory keep increasing as I am afraid one
> >it will use all memory in the machine, that will cause the entireallocated when
> >system hang. Unexpected server down is unacceptable to our client.
>
> The server itself is about 3Mb. With SS, the db cache gets
> the first client connection is made and is released when the lasttime
> connection logs out. With Classic the memory will increase each
> another client connection is made and will reduce each time aconnection is
> logged out. Each client pipe also adds to memory usage, not much,about
> 200 Kb per connection.a
>
> Still, I'm still very unclear about what your numbers are, or what
> difference of 432 between two numbers actually means. If it's Kbon
> Superserver, it might be just the difference between 8 connectionsand 10
> connections. If it's Mb on Classic, it might be the differencebetween one
> connection and two connections with a misconfigured db cache. Whoknows?
>cache
> If you have Forced Writes off, then at times you will have some OS
> holding the commits that are waiting to be flushed...without moreadequate
> information, who knows?
>
> heLen