ORA-27101: Shared Memory Realm Does Not Exist

ORA-27101: Shared Memory Realm Does Not Exist

Error Overview

Error Text: ORA-27101: shared memory realm does not exist

ORA-27101 means the Oracle client process cannot find the shared memory segment (SGA) for the target instance. The instance is likely not running, the ORACLE_SID environment variable points to a non-running instance, or the connection is going to a different host than expected.

Common Causes

Instance Not Started

• Database not started after server reboot
• Instance crashed
• Background processes died
• Listener registered service but instance is down

Wrong ORACLE_SID

• ORACLE_SID not set in shell
• ORACLE_SID points to different instance
• Multiple Oracle homes with different SIDs
• oraenv script returned wrong SID

Connection Routing

• TNS entry forces local bypass via (SERVER=DEDICATED) to non-existent local instance
• BEQ (bequeath) protocol used instead of network listener
• Listener not registered or wrong service name

Shared Memory Cleanup Issues

• Previous shutdown left stale memory but no instance
• IPC limits prevented SGA allocation
• Shared memory removed by ipcrm

Diagnostic Steps

Check Instance Status

# Verify ORACLE_SID
echo $ORACLE_SID

# Check if instance is running
ps -ef | grep "ora_pmon_$ORACLE_SID" | grep -v grep

# All Oracle instances on host
ps -ef | grep ora_pmon | grep -v grep | awk '{print $NF}' | sed 's/ora_pmon_//'

Verify Environment

# Set environment correctly
. oraenv
ORACLE_SID = [PROD] ?

# Or manual
export ORACLE_SID=PROD
export ORACLE_HOME=/u01/app/oracle/product/19c/dbhome_1
export PATH=$ORACLE_HOME/bin:$PATH

Test Local Connection

# Bypass listener with local connection
sqlplus / as sysdba

# If this fails with ORA-27101, instance is genuinely not running

Check Shared Memory

# View Oracle shared memory segments
ipcs -m | grep oracle

# Show kernel limits
sysctl -a | grep -E "shmmax|shmall|shmmni"

# Check current usage
cat /proc/sys/kernel/shmmax

Check Listener Status

# Listener health
lsnrctl status

# Look for service name registered
lsnrctl services

# Confirm instance is registered with listener

Resolution Steps

1. Start the Instance

# Set environment
. oraenv  # Choose correct SID

# Start
sqlplus / as sysdba
SQL> STARTUP;

If startup fails, check alert log:

adrci
adrci> show alert -tail 100

2. Set Correct ORACLE_SID

# Identify available SIDs
cat /etc/oratab

# Set the correct one
export ORACLE_SID=PROD

# Permanent in user profile
echo 'export ORACLE_SID=PROD' >> ~/.bash_profile
echo 'export ORACLE_HOME=/u01/app/oracle/product/19c/dbhome_1' >> ~/.bash_profile
echo 'export PATH=$ORACLE_HOME/bin:$PATH' >> ~/.bash_profile

3. Connect via Listener Instead

# Use TNS service name to route through listener
sqlplus user/pass@PROD

# Verify TNS entry
tnsping PROD

4. Clean Up Stale IPC

# Only after confirming no instance running for that SID
# List Oracle shared memory
ipcs -m | grep oracle

# Remove specific segment (CAUTION)
ipcrm -m <shmid>

# Remove semaphores
ipcs -s | grep oracle
ipcrm -s <semid>

# Better: use sysresv to validate before removing
sysresv -i -l PROD

5. Fix Shared Memory Limits

If startup fails with ORA-27101 plus ORA-04031 or ORA-27123:

# Edit /etc/sysctl.conf
kernel.shmmax = 17179869184    # 16GB or larger
kernel.shmall = 4194304        # in pages of 4KB
kernel.shmmni = 4096

# Apply
sysctl -p

# For larger SGAs, use HugePages
vm.nr_hugepages = 2048

6. Restart Listener

# Stop and start listener
lsnrctl stop
lsnrctl start

# Force instance registration
sqlplus / as sysdba
SQL> ALTER SYSTEM REGISTER;

# Verify
lsnrctl services

Common Scenarios

Scenario 1: After Server Reboot

$ sqlplus / as sysdba
ERROR:
ORA-01034: ORACLE not available
ORA-27101: shared memory realm does not exist
Linux-x86_64 Error: 2: No such file or directory

Fix: Database didn’t auto-start. Either run dbstart or set up systemd service.

Scenario 2: Wrong ORACLE_SID

ORACLE_SID=PROD set, but instance running is TEST.

Fix: export ORACLE_SID=TEST and retry.

Scenario 3: Listener Stale Registration

TNSping works, but connection returns ORA-27101.

Fix: Listener has stale entry for crashed instance. Restart listener.

Sample Output

$ echo $ORACLE_SID
PROD

$ sqlplus / as sysdba
ERROR:
ORA-01034: ORACLE not available
ORA-27101: shared memory realm does not exist
Linux-x86_64 Error: 2: No such file or directory
Process ID: 0
Session ID: 0 Serial number: 0

$ ps -ef | grep ora_pmon
oracle  12345  1  0 May08 ?  00:00:00 ora_pmon_TEST
# Only TEST is running, not PROD

# Option A: switch to TEST
$ export ORACLE_SID=TEST
$ sqlplus / as sysdba
SQL>  -- connected to TEST

# Option B: start PROD
$ export ORACLE_SID=PROD
$ sqlplus / as sysdba
SQL> STARTUP;
ORACLE instance started.
Database mounted.
Database opened.

Prevention Strategies

Auto-Start Configuration

# /etc/oratab - mark databases for auto-start
PROD:/u01/app/oracle/product/19c/dbhome_1:Y
TEST:/u01/app/oracle/product/19c/dbhome_1:N

# Use dbstart on boot
# Add to /etc/rc.d/rc.local
su - oracle -c "/u01/app/oracle/product/19c/dbhome_1/bin/dbstart $ORACLE_HOME"

systemd Service (Recommended)

/etc/systemd/system/oracle-db.service

[Unit]
Description=Oracle Database PROD
After=network.target

[Service]
Type=forking
RemainAfterExit=yes
User=oracle
Group=oinstall
ExecStart=/u01/app/oracle/product/19c/dbhome_1/bin/dbstart /u01/app/oracle/product/19c/dbhome_1
ExecStop=/u01/app/oracle/product/19c/dbhome_1/bin/dbshut /u01/app/oracle/product/19c/dbhome_1

[Install]
WantedBy=multi-user.target

systemctl enable oracle-db
systemctl start oracle-db

Environment Setup Discipline

/home/oracle/.bash_profile

. /etc/oratab.profile  # source standard env

# Helper functions
function go() {
    export ORACLE_SID=$1
    . oraenv <<< $1
}

# Usage: go PROD

Health Monitoring

#!/bin/bash
# Check all instances in /etc/oratab
while IFS=: read -r sid home auto_start rest; do
    [[ "$sid" =~ ^# ]] && continue
    [[ -z "$sid" ]] && continue

    if [ "$auto_start" = "Y" ]; then
        if ! pgrep -f "ora_pmon_$sid" > /dev/null; then
            echo "DOWN: $sid"
            mail -s "Oracle $sid not running" [email protected] < /dev/null
        fi
    fi
done < /etc/oratab

Document Per-Host Inventory

• Maintain a CMDB/wiki listing all instances per host
• Include SID, version, listener port, services
• Update on every install/decommission

Related Errors

• ORA-01034: ORACLE not available
• ORA-01089: Immediate shutdown in progress
• ORA-27102: Out of memory
• ORA-27123: Unable to attach to shared memory
• ORA-27125: Unable to create shared memory segment

Troubleshooting Checklist

• Verify ORACLE_SID matches running instance
• Check for ora_pmon_$SID process
• Confirm ORACLE_HOME is set correctly
• Try local connection (sqlplus / as sysdba) before TNS
• Check listener has correct service registered
• Inspect alert log for startup failures
• Validate kernel parameters for SGA size
• Configure auto-start to prevent recurrence