| Subject | Re: [ib-support] Firebird makes /tmp/core file, and it is HUGE |
|---|---|
| Author | Scott Taylor |
| Post date | 2002-08-02T15:06:38Z |
At 11:42 PM 01/08/2002, you wrote:
makes a lot of difference to your problem, although, you should update to
the stable release of RedHat 7.2 (not 7.3) for production servers.
tmp files which are usually created for searches. Check your
/opt/interbase/interbase.log for hints about bad client connections, i.e.:
"Super Server/main: Bad client socket, send() resulted in SIGPIPE, caught
by server
client exited improperly or crashed ????"
Doesn't seem to always clean up nicely.
I run a cron job that shuts down the FB/IB server and cleans out /tmp right
after the Sunday morning backup before anyone comes in (which may not
always be possible for some servers).
<rant>
I've caught users killing apps the Windows HarshWay (Ctrl+Alt+Del End-Task)
or just shutting off the PC while programs are running. Someone apparently
told them that was a safe way to shut down Windows and it is quicker so
they should do it all the time. Eventually the /tmp directory gets filled
up and FB/IB can not create a new sort file and doesn't like that.
</rant>
sorting.
/tmp a separate partition, and make it large (comparably). Depending on
how many users and how many different DB's are going to be accessed at the
same time, I use 20MB per user on-line * number of .gdb's that may be
accessed by the on-line users. Should be plenty.
One time I had /tmp as a directory off / instead of a separate partition,
that was just nasty, but the problem obvious.
HTH.
Scott.
>I have FirebirdSS-1.0.0.796 running on RedHat Linux 7.1.You are up-to-date with Firebird, but not running a stable OS, not that it
makes a lot of difference to your problem, although, you should update to
the stable release of RedHat 7.2 (not 7.3) for production servers.
>Every week or so I end up with a huge /tmp/core file, presumably some sortYour application needs to close the user out of FB/IB to clear off stale
>of core memory dump.
tmp files which are usually created for searches. Check your
/opt/interbase/interbase.log for hints about bad client connections, i.e.:
"Super Server/main: Bad client socket, send() resulted in SIGPIPE, caught
by server
client exited improperly or crashed ????"
Doesn't seem to always clean up nicely.
I run a cron job that shuts down the FB/IB server and cleans out /tmp right
after the Sunday morning backup before anyone comes in (which may not
always be possible for some servers).
<rant>
I've caught users killing apps the Windows HarshWay (Ctrl+Alt+Del End-Task)
or just shutting off the PC while programs are running. Someone apparently
told them that was a safe way to shut down Windows and it is quicker so
they should do it all the time. Eventually the /tmp directory gets filled
up and FB/IB can not create a new sort file and doesn't like that.
</rant>
>I have managed to reduce the size of the /tmp/core file by changing anThat's not a good idea, your FB server needs to create large files for
>entry in /etc/profile: ulimit -S -c 10000, so at least it no longer takes
>the database server down due to lack of drive space.
sorting.
>I would, however, prefer a solution that nips the problem in the bud.When you build your new server with the stable release, RedHat 7.2, make
/tmp a separate partition, and make it large (comparably). Depending on
how many users and how many different DB's are going to be accessed at the
same time, I use 20MB per user on-line * number of .gdb's that may be
accessed by the on-line users. Should be plenty.
One time I had /tmp as a directory off / instead of a separate partition,
that was just nasty, but the problem obvious.
HTH.
Scott.