SUMMARY
Oracle backups fail due to local mounted directory (/usr/bp/mnt/
ISSUE
Oracle backups fail due to local mounted directory (/usr/bp/mnt/<oracle_db_name>/<backup_type>.task) cannot be found. For Instance, customer has an Oracle database called 'test', and we are running a full backup on that database. Error in rae logs would indicate (note: n=the backup number that failed):
[LOG4] pghelper.c:1861: Executing query with text: SELECT status FROM bp.jobs WHERE job_no = n
[LOG4] pghelper.c:1247: row 0, col 0(status): '512'
[LOG4] pghelper.c:1861: Executing query with text: SELECT status FROM bp.jobs WHERE job_no = n
[LOG4] pghelper.c:1247: row 0, col 0(status): '512'
[LOG4] [methods.c:134] Reading complete. Read 83 bytes
[LOG0] [methods.c:597] Command "(no description)" completed with exit code 1
[LOG0] [methods.c:604] Details:
Method: BP-Proto
Execute: cp -f /usr/bp/mnt/test/oracle_full.n.task /tmp/oracle_full.n.task
Output:
cp: cannot stat `/usr/bp/mnt/test/oracle_full.n.task': No such file or directory
RESOLUTION
Perform the following steps to add fsid=0 to the exportfs command in the /usr/bp/bpinit/rae.conf.d/utilities.conf file:
- Save current file:
cp /usr/bp/bpinit/rae.conf.d/utilities.conf /usr/bp/bpinit/rae.conf.d/utilities.conf.orig
- Open /usr/bp/bpinit/rae.conf.d/utilities.conf for editing:
vi /usr/bp/bpinit/rae.conf.d/utilities.conf
- Search for 'exportfs:
/exportfs
- Line before change:
<Execute>exportfs -o rw,all_squash,anonuid=0,anongid=0 ${ClientName}:${SharePath}</Execute>
- Line after change:
<Execute>exportfs -o rw,all_squash,anonuid=0,anongid=0,fsid=0 ${ClientName}:${SharePath}</Execute>
- Save the file:
:wq!
- Restart nfs service on Oracle if stale
- Run a full backup of the Oracle Database
CAUSE
The issue occurs because we are not cleanly unmounting the NFS share, and causing a stale mount. The fix will properly mount and unmount the NFS share used for Oracle database backups.