i am an oracle DBA having solaris 10. in database alert log I am receiving following messages
Non Critical error ORA-48181 caught while writing to trace file "/u01/app/oracle/diag/rdbms/mygrid/mygrid12/trace/mygrid12_m000_210009.trc"
Error Messages: SVr4 Error: 28 : No space left on device
writing to the above trace file is disabled for now on
Error: Direct connection failure with ASM
WARNING:ASM communication error : op 0 state 0x0 (15055)
and in /var/adm/messages.0 showing following
May 20 14:08 xxxx ufs: [ID 845546 kern.Notice] NOTICE:alloc: /u01: file system full
out put of df -ah is as follwos
/dev/dsk/c0t500xxxxxxd0s4 439G 381G 54G 88% /u01
all other filesystem have free space shown in df -ah. I can't write all of them
Nowhere is that file system which is full as all file systems are not 100 % full and have sufficient free space
There is only one /u01 partition which I have shown or may be being beginner I am not understanding you as I don't see any other /u01 . There is only one mount point /u01 which points to the machine hardisk
But it is also an OS warning, so the Operating System interfered, so maybe that parameter was set to unlimited and there were unusually big trace files and maybe also log files that were removed by Oracle after the fact?
Yes it is set to unlimited but when I see the size of adump, cdump or .trc files all of them are in mbs and they are all in /u01 filesystem so at this moment my instance is down and I can't even start it and receiving following error
ORA-09817 : write to audit file failed
SVR4 Error: 28: No space left on device
Additional_information: 9
I thought also at something of the sort, as it happened to me one long ago, did not reply yet as I am still trying to figure what it was... but remembered once managed to umount (was not easy as a file was still open but could not be found as later I believed it was deleted...) after a few fsck it came back to normal...
Ah OK, I understand, you still need a "cooked" filesystem to write trace and log files..
What happens when you execute:
fuser -c /u01
?
One of those processes may still have a file handle to a file on the filesystem that has been deleted and thus the diskspace is not freed. You can use lsof to find out which processes have deleted files.
When processes hold it busy then it cannot be umounted.
Examine the PIDs shown by the fuser. The trailing letters show the kind of access, e. g. a c means "current work directory", and o means a file is opened.
ps -fp 1679
ps -fp 1670
...
Then try to terminate them in a friendly way.
E. g. use a shutdown procedure given by Oracle, or in an accessing shell cd out of it.
but i am worried what happen when my /u01 dismounted as there are 2 x instances ruuning as part of 2 Node RAC,. NOte that there is 7.2 GB of datat which i can delete from /u01. it is a folder at /u01/source. this sorce folder conatain sources of oracle etc so no use to me but ist i wan to copy it to sother SAN NFS share but it doesn't, like it did copied but not all data . i used cp -r /u01/source /export/home/oracle/source
the size is 7.2 Gb but when i copy it to destinaition the size is few Mbs. this is 2nd issue now