Hang when adding/resizing datafile
miam
I have a problem with Oracle 8i 8.1.6 on Linux Redhat 8.0 when I use :
alter tablespace ... add datafile '...' size 2000M;
Or :
alter tablespace ... add datafile '...' size 100M; (works well)
then :
alter database datafile '...' resize 2000M;
In this 2 cases, I have no error, but the statement doesn't return
control.
The oracle process on the server takes more and more cpu time, but the
hdd does nothing and the file has its required size on the disk.
I can't kill the process (I have tried kill -15, kill -9, and every
signal)
The only way : I must reboot the server
But when I restart the server and the database,
If I was adding the datafile, it is not mounted.
If I was resizing the datafile, it is not resized.
I have this problem on 2 different machines of my customers
What is this ?
And Oracle bug ? A Linux bug ?
I have not find it on internet :((
Thanks for your answers...
Daniel Morgan
I've no idea ... but why would you be in Oracle 8.1.6, unsupported,
and one of the buggiest versions of the database ever sold, after the
release of so many newer and better versions? Upgrades are free ...
upgrade!
--
Daniel A. Morgan
University of Washington
damo...@x.washington.edu
(replace 'x' with 'u' to respond)
Joel Garry
How long did you give it? Oracle has to "format" the file, and that
can take a long time. 816 may have bugs, too, patch it up. So write
a loop that extends it 100M at a time if you are impatient. That
might pick up any "large files" type of problem, too. You probably
shouldn't use files near 2G with older versions anyways, see
http://metalink.oracle.com/metalink/plsql/ml2_documents.showFrameDocument?p_database_id=NOT&p_id=62427.1
jg
--
@home.com is bogus.
Not exactly "The Last Convertible:"
http://www.infoworld.com/article/04/08/16/HNendofalpha_1.html
miam
Sorry, I forgot to mention it...
The process runs during more than 24 hours, but I don't know exactly
When I look for it the following day, it is often dead
I don't think the link you give treats this problem, because the size I
request is below 2GB (2000MB and not 2048MB)
"Joel Garry" <joel-...@home.com> a écrit dans le message news:
91884734.04081...@posting.google.com...
Joel Garry
> Hello
>
> Sorry, I forgot to mention it...
> The process runs during more than 24 hours, but I don't know exactly
> When I look for it the following day, it is often dead
>
> I don't think the link you give treats this problem, because the size I
> request is below 2GB (2000MB and not 2048MB)
Hmmm, what kind of disks/filesystems do you have? Are you using a
journaled filesystem? What happens when you make a new tablespace?
What blocksize?
Just grasping at straws here.
I did used to see these sorts of things happen on O8.0 hp and sun
systems, but don't recall the answers. You haven't turned on Oracle
tracing, by chance? Multiple database writers? OPS? Have you
checked to see what is waiting?
The process you can't kill, what is it's parent? Does a ps show it as
defunct or zombie? Are all of the Oracle processes there?
jg
--
@home.com is bogus.
Urban legends: http://www.infoworld.com/reports/33SRmythsofit.html
miam
> journaled filesystem? What happens when you make a new tablespace?
> What blocksize?
The problem occurs also when creating a new tablespace (and also when
creating tablespaces of a new database)
db_block_size=8192
>
> Just grasping at straws here.
>
> I did used to see these sorts of things happen on O8.0 hp and sun
> systems, but don't recall the answers. You haven't turned on Oracle
> tracing, by chance? Multiple database writers? OPS? Have you
> checked to see what is waiting?
I know "alter session set sql_trace=true", but I don't think this is
what you mean
How can I turn on Oracle tracing (I can reproduce the problem when I
want)
I don't know what the process is waiting :((
> The process you can't kill, what is it's parent? Does a ps show it as
> defunct or zombie? Are all of the Oracle processes there?
If the command is launched from remote, ps shows it as child of the
process "1", as every else remote-client-connection oracle process
If the command is locally launched, the process is child of the local
dba(svrmgrl)/client(sqlplus) program
if "MYDB" is the name of the database, the process is named "oracleMYDB"
and its environment is "(LOCAL=NO) or begin with "(DESCRIPTION=(LOCAL=Y"
depending if I have connected from remote or not
all oracle processes seem to be there, and have a normal behaviour :
ora_pmon_MYDB
ora_dbw0_MYDB
ora_lgwr_MYDB
ora_ckpt_MYDB
ora_smon_MYDB
ora_reco_MYDB
If I shutdown abort the database, the only process remaining is the
"oracleMYDB"
It is not killable