Skip navigation

On Monday April the 20th, around 6:30PM, on an 11.2.0.4 database, I launched 2 imports with IMP (the dumps were from an old version of Oracle).

 

imp system FULL=Y file=/data/migr/SRC20172020.dmp log=/data/migr/200420SRC20172020.log FEEDBACK=1100101

RESUMABLE=Y RESUMABLE_NAME=IRIS1720 IGNORE=Y

 

imp system FULL=Y file=EDE_REAS.dmp log=200420EDE_REAS.log FEEDBACK=1100101 RESUMABLE=Y RESUMABLE_NAME=EDE_REAS IGNORE=Y

 

FEEDBACK=1100101 means that while IMP is importing tables, I will see a dot each time it has imported 1,100,101 lines (why this weird number? Because it allows my eyes to tell my brain more quickly how much is that "1000000", for which my eyes would go "how many zeroes is that?" and waste precious milliseconds -- you know, performance… ).

But what I want to talk about is the 2 RESUMABLE parameters.

Then I can see lots of RED LOG switches as the import progrresses:

Unfortunately, at 7:43Pm and 7:49PM those 2 imports running concurrently maxed out a filesystem and there was no more physical space for the tablespaces to autoextend.  This is where RESUMABLE=Y came into play.

Mon Apr 20 19:43:30 2020

statement in resumable session 'IRIS1720' was suspended due to

    ORA-01653: unable to extend table IRIS_201706.PARAM_REAS by 128 in tablespace IRIS_201706_DATA

Exactly 2 hours later, the 2 imports timed out:

And so on every 2 hours for the rest of the night.

Then at 7:13AM on Tuesday morning

Staff is back at work at 7AM, so supervsion must have reported the maxed-out filesystem and on-call staff extended its size.

 

Then both imports resumed so REDO LOG switches resumed like crazy from 7:13Am onwards:

Then unfortunately at 8:52AM:

 

I must say that is not what I expected. Since  the FS became full at 7:49PM, I expected IMP to wait until 9:49PM for it to be extended and automatically resume, or abort if the FS extension did not occur. But stangely the warning that "the resumable session XXX was timed out" occured every 2 hours for the rest of the night, until it aborted at 8:52AM despite the fact that it had resumed at 7:13AM -- weird.  I'd love it if someone coud explain that behavior in the comments below.  Feel free to contribute your 2 cents or leave any kind of comment.