Re: Problems with tempdbs with IDS 9.30 UC1 and RedHat 7.1
Posted in 2006
Topics: Installation, Setup & Upgrades, Storage & Space Management, Server Administration, Versions, Editions & End-of-Life
Hello, we use our ERP to change a order. We now removed the tempdbs from the onconfig file. Now the IDS write to the directory /tmp. The server creates two files with aprox. 40 MB in the /tmp directory. The buffers in the Onconfig are set to 600.000. It seems, IDS did not do any paging in memory but immediately write all temp data to filesystem. Best regards Andreas bozon schrieb: > What is the query? > > Andreas wrote: > > Hallo, > > > > we have problems with a new installed IDS 9.30 UC1 on a RedHat 7.1 > > system. > > > > If we do a query on our databases, we get more than 400.000 pages > > written and near the same number of pages readed from the tempdbs > > partition. All other page written on partitions with rootdbs and > > datadbs seem to be normal. > > > > We are using a HP DL 380 G4 with 2 CPUs and 3.6 GB of memory. The > > RedHat runs on a virtual machine (VMware 2.5.3). A HP MSA1000 is used > > as disk. > > > > What could we do? > > > > Best regards > > Andreas
Sounds like bad statistics or missing indexes...
or incorrect settings for OPTCOMPIND -->> oltp set it to 0...
the engine could be doing a hash join instead of an index lookup
which may explain the huge number of writes to temp.
i would suggest to try and get the most expensive qry
and use
set explain on;...yourqry.
run that on your old server and new server. compare them!!!
Superboer
Andreas schreef:
> Hello,
>
> we use our ERP to change a order.
>
> We now removed the tempdbs from the onconfig file. Now the IDS write to
> the directory /tmp. The server creates two files with aprox. 40 MB in
> the /tmp directory. The buffers in the Onconfig are set to 600.000.
>
> It seems, IDS did not do any paging in memory but immediately write all
> temp data to filesystem.
>
> Best regards
> Andreas
>
>
> bozon schrieb:
>
> > What is the query?
> >
> > Andreas wrote:
> > > Hallo,
> > >
> > > we have problems with a new installed IDS 9.30 UC1 on a RedHat 7.1
> > > system.
> > >
> > > If we do a query on our databases, we get more than 400.000 pages
> > > written and near the same number of pages readed from the tempdbs
> > > partition. All other page written on partitions with rootdbs and
> > > datadbs seem to be normal.
> > >
> > > We are using a HP DL 380 G4 with 2 CPUs and 3.6 GB of memory. The
> > > RedHat runs on a virtual machine (VMware 2.5.3). A HP MSA1000 is used
> > > as disk.
> > >
> > > What could we do?
> > >
> > > Best regards
> > > Andreas
Hello,
could it be an UPDATE STATISTICS problem. The database was imported
from the old server.
Best regards
Andreas
Superboer schrieb:
> Sounds like bad statistics or missing indexes...
> or incorrect settings for OPTCOMPIND -->> oltp set it to 0...
> the engine could be doing a hash join instead of an index lookup
> which may explain the huge number of writes to temp.
>
> i would suggest to try and get the most expensive qry
> and use
>
> set explain on;> ...yourqry.
>
> run that on your old server and new server. compare them!!!
>
>
> Superboer
>
> Andreas schreef:
>
> > Hello,
> >
> > we use our ERP to change a order.
> >
> > We now removed the tempdbs from the onconfig file. Now the IDS write to
> > the directory /tmp. The server creates two files with aprox. 40 MB in
> > the /tmp directory. The buffers in the Onconfig are set to 600.000.
> >
> > It seems, IDS did not do any paging in memory but immediately write all
> > temp data to filesystem.
> >
> > Best regards
> > Andreas
> >
> >
> > bozon schrieb:
> >
> > > What is the query?
> > >
> > > Andreas wrote:
> > > > Hallo,
> > > >
> > > > we have problems with a new installed IDS 9.30 UC1 on a RedHat 7.1
> > > > system.
> > > >
> > > > If we do a query on our databases, we get more than 400.000 pages
> > > > written and near the same number of pages readed from the tempdbs
> > > > partition. All other page written on partitions with rootdbs and
> > > > datadbs seem to be normal.
> > > >
> > > > We are using a HP DL 380 G4 with 2 CPUs and 3.6 GB of memory. The
> > > > RedHat runs on a virtual machine (VMware 2.5.3). A HP MSA1000 is used
> > > > as disk.
> > > >
> > > > What could we do?
> > > >
> > > > Best regards
> > > > Andreas