One of the more interesting Oracle failures is when the database can reach the MOUNT state but fails while opening. That distinction provides an important clue about where the problem is occurring.
The Situation
After an Oracle server problem, I needed to bring the database back online. I first verified the ASM infrastructure before attempting to open the database.
sqlplus / as sysdba
The instance progressed far enough to mount, but opening the database failed with a connection-related error such as:
ORA-03113: end-of-file on communication channel
Why MOUNT vs. OPEN Matters
NOMOUNT
↓
MOUNT
↓
OPEN
During NOMOUNT, Oracle starts the instance and reads initialization parameters. During MOUNT, Oracle reads the control files. During OPEN, Oracle accesses and validates database files and performs the work required to make the database available.
If STARTUP MOUNT succeeds but ALTER DATABASE OPEN fails, the problem is probably not simply an inability to start the Oracle instance. The investigation should focus on what Oracle is doing during the transition from MOUNT to OPEN.
Do Not Diagnose ORA-03113 by Itself
ORA-03113 often is not the actual root cause. It tells the client that communication with the server-side process ended unexpectedly. The more useful evidence is normally in Oracle's diagnostic files.
tail -100 alert_orcl.log
tail -f alert_orcl.log
I also check the newest trace files under the Oracle diagnostic destination:
cd $ORACLE_BASE/diag/rdbms/orcl/orcl/trace
ls -ltr | tail
The trace file may contain the actual Oracle or operating-system error that caused the server process to terminate.
Verify ASM Independently
Because this environment used ASM, I verified ASM separately:
sqlplus / as sysasm
SELECT name, state, type
FROM v$asm_diskgroup;
This separates two different questions: Is ASM available? and Can the database successfully open its files?
Test the Startup Phases Separately
STARTUP NOMOUNT;
ALTER DATABASE MOUNT;
ALTER DATABASE OPEN;
That turns a broad statement such as “the database won't start” into a much narrower observation: “the instance starts and the control files are readable, but the failure occurs while opening the database.”
Lesson Learned
Production troubleshooting is often about reducing the problem space one layer at a time: operating system → ASM/storage → Oracle instance → control files → database files → recovery/open.
At each layer, verify what is known to be working before moving to the next one. That approach avoids unnecessary restarts and prevents assumptions from driving the investigation.