Hi,
Thanks for the response !
> Howdy,
>
> On Tue, Jan 09, 2001 at 11:53:12AM +0800, [EMAIL PROTECTED] wrote:
> > I've got a linux server running MySQL 3.22.32 on kernel version
> > 2.4.0-test8.
>
> I'd be tentatively inclined to blame the problems on 2.4.0-test8. Do
> the problems still occur under (say) 2.2.18?
Upgrading the kernel is the next step. 2.4.0-test8 is a pretty standard
version here, and this is the only machine showing problems, but you're
right to question it.
> > The database server has a number of housekeeping scripts running
> > in the background (around half a dozen), all written in Perl, and
> > the database itself is updated from 4 linux machines elsewhere on
> > the network. One in particular is capable of generating between
> > 3000 and 9000 inserts into a table called CurrentTraffic every 5
> > minutes. Records are removed from the table as required by one of
> > the scripts running in the background.
>
> Sounds reasonable.
>
> > Periodically, one of the database tables gets corrupted. It's always
> > CurrentTraffic, and running isamchk against the table reports that
> > the delete list is corrupt. Yesterday, I inserted about 450 Mb
> > into the table, and found that it got screwed up. I don't know how
> > many records there were ("select count(*) from CurrentTraffic" ran
> > for over an hour before I gave up on it), but the table definition
> > isn't enormous - one char(15) and twelve int(10)'s - and
> > CurrentTraffic.ISD was about 530 Mb, so I'd guess at somewhere about
> > 9.5 million rows.
>
> As I understand it, a query like `select count(*) from CurrentTraffic'
> is a special case and should return immediately...? Unless the table
> is locked by a writer, of course.
I don't explicitly lock any of the tables during the bulk inserts .. wouldn't
doing so essentially make accesses to the table single-threaded? There's
plenty of room for performance issues already, so I'd rather not deliberately
build in a bottle-neck like that. I've thought of it of course ...
> Do you stop the database before running isamchk?
Yes.
> Are you hitting a 2GB limit at any stage? (Yeah, I know you're running
> 2.4, but ISTR 3.22 and isam tables have some intrinsic 2GB limits.)
>
> Do you have many indices? Is it possible the .ISM file is hitting the
> 2GB limit while the .ISD file is not?
The data file was 530 Mb, and the index file 260 Mb. 3 fields are indexed,
no primary key, just simple indexes.
> > I've read the documentation for MySQl, and paid attention to the
> > discussions here, but there's no mention of problems explicitly related
> > to the kind of setup that I've got here.
> >
> > Is there an issue with bulk inserts coming over a network whilst
> > (different) records are being deleted by local scripts? I can't
> > imagine it, but who knows?
> >
> > Any other ideas are, of course, welcome !!
>
> Have you tested 3.23.latest?
Not yet, no. I had felt (perhaps wrongly) that the latest supported
release was a better bet than using something from the development tree,
although I *do* understand that the MySQL development branch is perhaps
more stable than some commercial products !!
We're in the process of arranging comemrcial support, which also made me
look at 3.22.x in preference.
The Hooker
--
"Did the aliens forget to remove your anal probe?"
---------------------------------------------------------------------
Before posting, please check:
http://www.mysql.com/manual.php (the manual)
http://lists.mysql.com/ (the list archive)
To request this thread, e-mail <[EMAIL PROTECTED]>
To unsubscribe, e-mail <[EMAIL PROTECTED]>
Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
