Will the db crash due to missing the current log file?

[abongwa]

The following is quoted form Oracle doc.

"Note that the current log is the one LGWR is currently writing to. If a LGWR I/O fails, then LGWR terminates and the instance crashes. "

I did a test on a test db. I drop the current online log file while the database is up and running. The db did not crash. I can continue read and write to the database. So, why Oracle's doc says that?

[pando]

may be when you dropped the lgwr wasnt writing?

[thanigai]

Did you check the alert log file....
Log might have switched before you deleted it...Try to switch log file and keep LGWR busy it will crash....
It happened couple more times to us...

Any update please post ...
Thanks.
Thanigaivasan.

[pando]

well it never happen to me so may be interesting hearing from someone

try create bigger redo like 100MB or 200MB and see what happens if you drop one

[abongwa]

It just don't crash!!! Can somebody try it on your test db?

[thanigai]

Hi we ask our newly joined dba to play likes this in test database kept seperatly for that purpose..we got this there only.We will try crash db all means such a way that we no need to update resume while working in production db...by doing doing same mistake.
Thanigaivasan.

[abongwa]

I could not understand your English.

[jmodic]

Originally posted by abongwa
It just don't crash!!! Can somebody try it on your test db?

Don't be silly. If it doesn't crash, there might be reasons for this. Here are some of them:

a) You have deleted wrong file, the one that is not a redo log of this instance
b) You deleted inactive online redo log and logwritter has not yet come to a point when it needs to switch to that redo log. Try forcing log switch manualy couple of times and see what happens
c) You have multiplexed logfiles in redo groups and you deleted only one file from the group. In this case Oracle continues to write to the remaining member(s) of a group, so no need to instance crash

Why don't you try querying some V$ views to see what is happening (V$LOG, V$LOGFILE, ...)

[abongwa]

Why use "silly"?

I checked all the points you mentioned here. This is for proof of concept purpose. The db has three log group with only one member each. It's on archivelog mode. When I remove the current log, I can continue read and write to the db. I can switch log groups UNTIL it "recycle" back to the log that has been removed. Then, the db just hang. It won't crash. I can even shutdown it nicely. Then, when I tried to start it up. It can mount but not open.

I then use "alter database clear unarchived logfile group 1;"
The db go back to normal. This is different from what Oracle said in the Doc:
"If a LGWR I/O fails, then LGWR terminates and the instance crashes. "

Note, this is for proof of concept. It's like you have to test your parachute's functions before you go skydiving so that you can succeed in the first time and anytime without DOUBT.

[jmodic]

Originally posted by abongwa
Why use "silly"?

Because your questions didn't make any sence, your cry for help was just plain silly! And when thanigai tried to help you with his best intentions, you simply turned him down with "I could not understand your English."! No "Thank you, I appreciate your good will" or something like that, only "I could not understand your English", cold and presumptuous.

And it wouldn't hurt if you described your whole scenario in your first post, like you did just now (although some vital information are still missing). That way you would probably get more accurate answers in the first place, and would save some time to all of us who read or tried to help in this thread...