Re: 2 gig per raw space

Thuc Nguyen wrote: > > Hi everyone, > > I am with Online 5.03 on HP 9.02 > > My "Sr" DBA did not have full confidence in what I told him. > > I have seen the previous postings but did not save them. I need to prove > him that I am right by showing him your postings. > > Am I correct that there is limit of 2 gig per raw space? > > Am I correct that if we have 8 gigs of hard disk then we will have 4 two > gigs raw spaces? > > Example: > > /dev/ifmx/coris0 2 gigs > /dev/ifmx/coris1 2 gigs > /dev/ifmx/coris2 2 gigs > /dev/ifmx/coris3 2 gigs > > Am I correct that there is no limit of gigs each dbspace? Version 7.x supports up to 2048 Chunks, each with a size of 2GB. Since there is a limit of chunks you can store up to 4TB. Version 5.x has a smaller limit of chunks. This limit depends on the length of your chunk pathnames. Try to keep the pathnames short - use links ! > Am I correct that the fragment can work on Online 5.03? if so, how much > is limited per dbspace? Fragemnet ? What kind of fragment do you mean ? > BTW, I saw the log file. It said, something like "time out..page > cleaner" What does it mean? Is it running out of page cleaner? There > was four page cleaners in the tbconfig file. Do you have 4 different disks ? Why did you configure 4 page cleaners ? When a page cleaner dies, it had a lot to do and couldn't perform the neccessary writes in time. It looks like a strange behaviour when a cleaner dies, but in most cases this will improve the performance. Just for your info, my system is running since 230 days, two page cleaners for two different disks are installed, both cleaners are running. If you would install 4 cleaners for a single disk ( maybe with 4 chunks ), the different cleaners would try to write to your disk concurrently. Try to simulate the situation, then you will see why the cleaners want to die. Try to READ 40MB with only one cleaner ( we will use dd instead ) time dd if=/dev/ifmx/coris0 of=/dev/null bs=2048 count=20000 Note the real time, and continue with the next simulation. There we will use a script: -----------------script------------ : dd if=/dev/ifnx/coris0 of=/dev/null bs=2048 count=5000 & dd if=/dev/ifmx/coris0 of=/dev/null bs=2048 skip=5000 count=5000 & dd if=/dev/ifmx/coris0 of=/dev/null bs=2048 skip=10000 count=5000 & dd if=/dev/ifmx/coris0 of=/dev/null bs=2048 skip=15000 count=5000 & wait ----------------------------------- chmod 755 ./script time ./script If this example will take a little longer :;) then you will understand that it is sometimes better to use only one process per disk. When a page cleaner needs a lot of time to perform it's output operation, the page cleaner thinks of this situation. > > I understand the extent size method but how can I determine how much I > need to add to the next size? If you understood the first extent size, then you will rarely need next extents. In this case set your next size to about 30% of your first extent size. > Thanks in advance for your reply. > > Thuc Bye Stefan