Data block in cache

hi
i have two questions...

1) when we try to retrieve data from a table...first cache is checked and then if not found
then it goes to disk for physicla read...right?
is this behaviour good that even if we need just a single row from a data block..whole block
or 4k or 8k is cached in the memory.....am i saying right??? is this the behaviour???

2) when a user performs a commit...the changed blocks are written permanently to disk at that
time or writing to disk happens just on check point ???

Regasrds

1.)
yes
yes
yes

2.)
At the time of commit the changes are permanently written to disk. But they are not necesserily written into database files at that time, they might only be written into redolog files.

Quote:

---

Originally posted by jmodic
1.)
yes
yes
yes

2.)
At the time of commit the changes are permanently written to disk. But they are not necesserily written into database files at that time, they might only be written into redolog files.

---

JMODIC,
I'm not sure about my question, and I need some enlightenment on this.
So, if the user perform the commit statement, the actual data isn't yet saved. It will only be written after the checkpoint timeout. And if the data is not written all the time in database files they might only be written in redolog files isn't it dangerous when a redolog files itself are corrupted(redo logs all in one disk)? If you're going to recover your database, and after then you use RESETLOGS to open the db since the redologs was damaged, then data could be lost.
Or did I miss something again.
thanks,
rey:confused:

Quote:

---

Originally posted by reydp
And if the data is not written all the time in database files they might only be written in redolog files isn't it dangerous when a redolog files itself are corrupted(redo logs all in one disk)? :confused:

---

Thatswhy you are told again and again to keep the mulitple members of the redo log groups on separate disks.

Quote:

---

Originally posted by SANJAY_G
Thatswhy you are told again and again to keep the mulitple members of the redo log groups on separate disks.

---

Thanks,
I'm not speaking not just for my self SANJAY_G, but i'm sure you're aware that the default installations of oracle uses the same drive for redologs. So, if I were new in DBA, I wouldn't know that my DB can still loose its data despite the commit statement, even if I can set the ARCHIVELOG on.

I'm sure, there is a very rear change of data loss because, that's what ORACLE's claims including us. Specially if the DB is very well setup. But my only point is, if there is a chance of a data loss, then this could be one reason of it.

I'm on trying to make a point here, which I only follow to Jmodic's opinion.:)

Yes, there is possibility of loosing data if the all online redo logs are lost before they are archived.

Quote:

---

Originally posted by SANJAY_G
Yes, there is possibility of loosing data if the all online redo logs are lost before they are archived.

---

many thanks SANJAY_G,
This is all that I want to hear. :D