Subject | Re: Firebird server blocks |
---|---|
Author | derryck.welas |
Post date | 2009-12-17T11:41:40Z |
Hi,
dont think it's a bug (i run 2.1.3 classic on several 8 core boxes Linux and Win db size up to 50GB) , more like config but you need to give more information .
Can you post gstat -h info ?
lockHashslots check if you have enough of those with "fb_lock_print".
Also there is the possibility that your client connection (lib) has a problem.
Have you tried connecting to the database from an other machine with a different connection client (java: squirel) or c++:flamerobin)
What i find strange is that you are using Superserver on an 8 core machine while Classic would be better to use the cores in parallel.
Cheers,
Derryck
dont think it's a bug (i run 2.1.3 classic on several 8 core boxes Linux and Win db size up to 50GB) , more like config but you need to give more information .
Can you post gstat -h info ?
lockHashslots check if you have enough of those with "fb_lock_print".
Also there is the possibility that your client connection (lib) has a problem.
Have you tried connecting to the database from an other machine with a different connection client (java: squirel) or c++:flamerobin)
What i find strange is that you are using Superserver on an 8 core machine while Classic would be better to use the cores in parallel.
Cheers,
Derryck
--- In firebird-support@yahoogroups.com, Pasi Parkkinen <pasi.parkkinen@...> wrote:
>
> Hi,
>
> We are running currently a database with FirebirdSS-2.1.2.18118-0 at Linux box with 8 cpu cores and 8G RAM.
>
> Server platform: 2.6.18-164.6.1.el5 #1 SMP Tue Nov 3 16:12:36 EST 2009 x86_64 x86_64 x86_64 GNU/Linux (CentOS 5.4)
>
> Database has quite a lot of traffic (intensive inserts and updates), file size is at the moment approx 1.6G
>
> Occasionally we are getting situations where Firebird is not available anymore to respond to requests, I think there might be a deadlock or similar at the database. If I restart all applications which are using the database it will not help. Only trick which helps is to restart the Firebird instance. After that everything works fine.
>
> It's a production box, so I haven't had a possibility (yet) for searching locks with fb_lock_print -tool - on the other hand I'm quite inexperienced with that, whatsoever my question for you is that is this a bug when we can put Firebird to a state that it has to be restarted? If this is all about deadlocks - is there any way to release them without dbengine restart, which I believe can cause data corruption??
>
> BTW. This has been seen at least from 2.0.3 to 2.1.2 (In some other environments I think also with 2.1.3, but not 100% sure). A Debug version would perhaps be nice, but it's not possible at this production environment: At firebird.log -file there's nothing ...
>
> Thanks for you help!
>
> -Pasi
>
>
> [Non-text portions of this message have been removed]
>