| Subject | Re: [firebird-support] server hang ups |
|---|---|
| Author | Helen Borrie |
| Post date | 2008-03-02T21:59:37Z |
At 07:13 AM 3/03/2008, you wrote:
You have the option of reverting to the Background method alone, which would at least take you back the GC conditions that you had with Fb 1.5. Or, if you have plenty of machine resources, you could use Classic instead, where Cooperative is the only method used.
However, neither of these options is going to correct the problem that *causes* the build-up of garbage. That problem lies in your client code.
You should also provide *honest* information about how often you perform gbak on the database without the -g switch and/or sweeping. Meaning, when people have databases with severe garbage collection problems, they tell you what the housekeeping routines are *meant* to be, not what they actually DO! ;-)
If poor GC *is* the basis of your troubles then this list can help you with the forensics and explain what admin measures you can take to reduce the effects. We can explain what the application code needs to do to avoid those troubles recurring.
However, for advice about altering your PHP scripts and effecting a solution, you would need to use the firebird-php list. I'm sure the firebird-php community would welcome the participation of a new face with a long history of using PHP for access to Firebird, as well.
[ If this is an O2 service provider we are looking at here (just a wild guess!) then you're not the only one who could be very keen to get a perfect solution to what looks like a very long-standing problem. ]
./heLen
>Hi all,This smells like a garbage collection (and transaction management) problem.
>
>i use firebird or interbase from version 4. All that time there was
>small or bigger issues with it, but nothing so serious to blame it.
>
>Till now of course :)
>
>I'm small hosting provider. We host near 200 websites on server.
>Majority of them uses firebird of course. As we use gentoo we used
>firebird from official repository. Client is PHP on apache. Again base
>gentoo versions.
>
>Till last weekend we used version 1.5.3. Everything worked near
>perfect. Except one think. Firebird used hang every 10 days but later
>every 24 hours.
>It simple stop responding to connect or react on any api function. ItThis smells like an exacerbated garbage collection problem.
>simple locks out. There is no load from firebird on server and it
>looks its in Sl state (sleep multithreads.) Not respondindg to TERM
>only to KILL
>Last week I upgraded to version 2.0.3 issue is even worse.Have you studied the release notes? In particular, search for "garbage collection". Fb 2 offers more options for GC on Superserver and installs with two GC methods combined. Without more information, e.g., the DB header statistics and whether you upgraded the database to ODS 11, the obvious deduction is that you have a long-term GC problem and cooperative GC is making it worse.
You have the option of reverting to the Background method alone, which would at least take you back the GC conditions that you had with Fb 1.5. Or, if you have plenty of machine resources, you could use Classic instead, where Cooperative is the only method used.
However, neither of these options is going to correct the problem that *causes* the build-up of garbage. That problem lies in your client code.
>so i use:You will have the same problem with PostGres. Worse, actually, because with PG you have to take your database offline to deal with garbage overload.
>firebird superserver 2.0.3.12981-r5 (portage version)
>php-5.2.5_p20080206-r3 (portage version)
>32bit gentoo-sources kernel 2.6.24-gentoo-r3
>CFLAGS="-O2 -march=i686 -pipe"
>CHOST="i686-pc-linux-gnu"
>
>Daily aprox. 1 milion of connects to server (hits), i do not use
>persistent connection as it had no impact on issue.
>
>And here is question:
>Can anyone help me to solve this issues? If not i'll be forced to
>replace firebird by something else (sqlite and postgres). But as i
>said i use firebird for long time and i like it.
>So If there is anyone who can point me where problem is i'll be veryThere's no need to send the database, nor anything to learn from it. You need to do your own forensics. But, if you want help with this you need to provide more information. The output from gstat -h would be a good starting place, since it will tell us what the ODS of the database is, besides providing some essential clues about how your application code is managing transactions.
>happpy. I can attach gdb on hanged firebird. But i cannot use strace
>ot run firebird in gdb as this issue happen on production server. I
>can provide any aditional information on request.
You should also provide *honest* information about how often you perform gbak on the database without the -g switch and/or sweeping. Meaning, when people have databases with severe garbage collection problems, they tell you what the housekeeping routines are *meant* to be, not what they actually DO! ;-)
If poor GC *is* the basis of your troubles then this list can help you with the forensics and explain what admin measures you can take to reduce the effects. We can explain what the application code needs to do to avoid those troubles recurring.
However, for advice about altering your PHP scripts and effecting a solution, you would need to use the firebird-php list. I'm sure the firebird-php community would welcome the participation of a new face with a long history of using PHP for access to Firebird, as well.
[ If this is an O2 service provider we are looking at here (just a wild guess!) then you're not the only one who could be very keen to get a perfect solution to what looks like a very long-standing problem. ]
./heLen