PowerShell scripts for setting up Oracle 21c/XE with ROA application:
- Automated tablespace, user creation and imports
- sqlnet.ora config for Instant Client 11g/ODBC compatibility
- Oracle 21c read-only Home path handling (homes/OraDB21Home1)
- Listener restart + 10G password verifier for legacy auth
- Tested on VM 302 with CONTAFIN_ORACLE schema import
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add config/sqlnet.ora with ALLOWED_LOGON_VERSION=8 for old client support
- Add scripts/fix-sqlnet.sh startup script to persist config across container restarts
- Update README with ORA-28040 troubleshooting, ODBC connection params, and deployment instructions
- Fix SID description: Oracle 18c has PDB (XEPDB1), not non-CDB
- Update container recreation instructions with startup scripts volume
Resolves ORA-28040: No matching authentication protocol when connecting
from Windows ODBC with Oracle Instant Client 11.2
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Rename proxmox/claude-agent/ to proxmox/lxc171-claude-agent/
- Move scripts to scripts/ subdirectory
- Add complete installation guide for new LXC from scratch
- Update proxmox/README.md with LXC 171 documentation and navigation
- Add LXC 171 to containers table
- Remove .serena/project.yml
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Move oracle/migration-scripts/ to proxmox/lxc108-oracle/migration/
- Move oracle/roa/ and oracle/roa-romconstruct/ to proxmox/lxc108-oracle/sql/
- Move oracle/standby-server-scripts/ to proxmox/vm109-windows-dr/
- Move chatbot/ to proxmox/lxc104-flowise/
- Update proxmox/README.md with new structure and navigation
- Update all documentation with correct directory references
- Remove unused input/claude-agent-sdk/ files
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Create cluster/ for Proxmox cluster infrastructure (SSH guide, HA monitor, UPS)
- Create lxc108-oracle/ for Oracle Database documentation and scripts
- Create vm201-windows/ for Windows 11 VM docs and SSL certificate scripts
- Add SSL certificate monitoring scripts (check-ssl-certificates.ps1, monitor-ssl-certificates.sh)
- Remove archived VM107 references (decommissioned)
- Update all cross-references between files
- Update main README.md with new structure and navigation
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace FORALL bulk operations with FOR loops to avoid PLS-00436 error
on Oracle 10.2.0.5. The older Oracle version does not support referencing
record fields from collection in FORALL statements.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix permission denied on log files (chown nut:nut)
- Fix upssched.conf permissions (root:nut)
- Add sudo for perl to allow PVE::Notify from user nut
- Add periodic battery status emails every minute when on battery
- Add charging status emails at 5, 10, 30 min after power restore
- Remove diacritics from all notification messages
- Update documentation with sudo and permissions setup
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add email notifications via PVE::Notify for all UPS events:
- ONBATT: when UPS switches to battery
- ONLINE: when power is restored
- LOWBATT: critical battery level
- SHUTDOWN_START/NODE/PRIMARY: during cluster shutdown
- COMMBAD: communication lost with UPS
- Add automatic UPS shutdown command after cluster shutdown
(protects against power surge when power returns)
- Update upssched.conf with ONLINE handler and immediate ONBATT notification
- Add notification templates for HTML and text emails
- Update documentation with new features and timer configuration
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Document Flowise and ngrok configuration on LXC 104, including
troubleshooting steps for CORS and version issues.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Storage Configuration improvements:
- Add "Noduri" column showing which nodes have access to each storage
- Clarify that 'local' is separate on each node (non-shared)
- Clarify that 'local-zfs' is shared across pvemini, pve1, pveelite
- Clarify that 'backup' is only on pvemini (10.0.20.201)
- Add detailed explanations for each storage type
- Add storage paths section with important locations
Node name corrections:
- Fix node name: pve2 → pveelite (correct cluster name)
- Update all references across proxmox-ssh-guide.md and README.md
- Add node descriptions in tables for clarity
Benefits:
- Users now know exactly which storage is available on which nodes
- Clear distinction between shared and non-shared storage
- Correct node naming throughout documentation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Move oracle/CONEXIUNI-ORACLE.md → proxmox/oracle-database-lxc108.md
- Create proxmox/README.md as documentation index
- Update proxmox-ssh-guide.md:
* Remove VM 107 references (decommissioned)
* Update LXC and VM tables with IP addresses
* Add IP address map for all services
* Simplify Oracle section (detailed info in oracle-database-lxc108.md)
* Update backup job configuration
Benefits:
- All infrastructure docs in proxmox/ directory
- Clear separation: general Proxmox (proxmox-ssh-guide.md) vs Oracle-specific (oracle-database-lxc108.md)
- No duplicate information between files
- Easy navigation with README.md index
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Problem: DR weekly test failed with ORA-00600 [kcbzib_kcrsds_1] when executed
via cron, but succeeded when run manually. Error occurred during "ALTER DATABASE
OPEN RESETLOGS" step after successful restore and recovery.
Root Cause Analysis:
- Manual test (12:09): Undo initialization = 0ms, no errors
- Cron test (10:45): Undo initialization = 2735ms, ORA-00600 crash
- Alert log showed: "Undo initialization recovery: err:600"
- Oracle instance was in inconsistent state from previous run
The cleanup_database.ps1 script had an "optimization" that preserved the
running Oracle service to "save ~30s startup time". This left the service
in an inconsistent state between test runs, causing Oracle to crash when
attempting to open the database with RESETLOGS.
Solution:
Modified cleanup_database.ps1 to ALWAYS stop Oracle service completely:
1. SHUTDOWN ABORT the instance (not just when /AFTER flag)
2. Stop-Service OracleServiceROA (force clean state)
3. Kill remaining oracle processes
4. Service starts fresh during restore (clean Undo initialization)
Changes:
- Removed if/else branch that skipped shutdown before restore
- Always perform full shutdown regardless of /AFTER parameter
- Updated messages to reflect clean state approach
- Added explanation: "This ensures no state inconsistencies (prevents ORA-00600)"
Testing: Manual test confirmed clean 0ms Undo initialization after fix.
Related: Works in conjunction with weekly-dr-test-proxmox.sh PATH fix (commit 34f91ba)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Problem: The weekly DR test script worked when run manually but failed
when executed via cron with "Failed to start VM 109" error at 0 seconds.
Cause: Cron jobs run with a minimal PATH that doesn't include /usr/sbin
where Proxmox commands (qm, pvesh, etc.) are located. Manual execution
had the full PATH including /usr/sbin.
Solution: Added explicit PATH export at the start of the script to ensure
all required system binaries are accessible regardless of execution context.
Testing: Successfully verified with cron test at 11:32 - VM started properly,
restore process completed normally.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Fixed two critical issues with HA monitoring:
1. False positive quorum errors - corosync-quorumtool not in cron PATH
2. Unwanted cron emails from PVE::Notify INFO messages to STDERR
Changes:
- Set proper PATH including /usr/sbin for corosync-quorumtool
- Split notification code: verbose shows all, non-verbose redirects STDERR to /dev/null
- Prevents cron from sending duplicate notification emails
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changed -mtime logic from +$RetentionDays to +($RetentionDays - 1) to correctly implement 2-day retention. Previously kept 3 days (today + 2 previous), now keeps exactly 2 days (today + yesterday).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Problem: Backups accumulated on DR (73 files, 4 days) instead of keeping only 2 days
- transfer_incremental.ps1 had no cleanup function (ran 2x/day without cleanup)
- transfer_to_dr.ps1 cleanup had poor logging
- oracle-backup-monitor-proxmox.sh couldn't detect new L0/L1 backup format
Changes:
- Add cleanup to transfer_incremental.ps1 (delete backups older than 2 days)
- Improve cleanup logging in transfer_to_dr.ps1 (shows count before/after)
- Update oracle-backup-monitor-proxmox.sh to detect both naming conventions:
* Old: *FULL*.BKP, *INCR*.BKP
* New: L0_*.BKP (Level 0), L1_*.BKP (Level 1)
- Remove temporary files from /input/ directory
Result: Monitor now correctly reports backup age, cleanup runs after each transfer
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Replace complex SSH+PowerShell query with simple log file parsing
- rman_restore_from_zero.ps1 already verifies and outputs database status
- Parse 'OPEN_MODE: READ WRITE' and 'TABLES: <count>' from LOG_FILE
- Fixes issue where successful restore was reported as FAILED
- More reliable: avoids SSH escaping issues with Select-String -Quiet
Root cause: SSH+PowerShell+sqlplus+Select-String chain was too fragile and
returned empty/false even when database was successfully opened (42625 tables).
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
- Add full path to FORMAT in rman_backup.txt and rman_backup_incremental.txt
- Files now stored in C:\Users\oracle\recovery_area\ROA\autobackup- Fixes issue where backups were created in ORACLE_HOME\DATABASE instead of recovery area
- Ensures transfer_to_dr.ps1 can find and transfer all backups correctly
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
- Add FORMAT to rman_backup.txt: L0_*, ARC_*, SPFILE_*, CF_*
- Add FORMAT to rman_backup_incremental.txt: L1_*, ARC_*, SPFILE_*, CF_*
- Update rman_restore_from_zero.ps1 TestMode to select files by naming convention
- Select only latest L0 backup set + all L1 incrementals/archives (faster DR tests)
- Backward compatible with old autobackup naming (fallback to copy all)
- Fixes missing datafiles issue (previously only copied 8 files, now copies full backup set)
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Critical fix based on user analysis:
PROBLEM:
Cleanup is called in 2 contexts with different requirements:
1. BEFORE restore (from rman_restore): Should NOT shutdown
2. AFTER restore (from weekly-test): MUST shutdown to delete files
USER INSIGHT:
"Why shutdown if restore will clean anyway? But AFTER restore,
you MUST shutdown to release file locks for deletion!"
SOLUTION:
Add /AFTER parameter to cleanup_database.ps1:
WITHOUT /AFTER (before restore):
- Skip SHUTDOWN ABORT
- Skip Stop-Service
- Leave service in current state (running/stopped)
- Files CAN be deleted (no lock before restore)
- Optimization: If service running → restore saves ~30s
WITH /AFTER (after restore):
- SHUTDOWN ABORT (stop instance)
- Stop-Service (release file locks)
- REQUIRED for file deletion after restore
- Files are locked by active instance/service
CALL SITES:
1. rman_restore: cleanup_database.ps1 /SILENT (no /AFTER)
2. weekly-test: cleanup_database.ps1 /SILENT /AFTER (with /AFTER)
FLOW OPTIMIZATION:
Test 1: Service stopped → start(30s) → restore → cleanup /AFTER
Test 2: Service stopped → start(30s) → restore → cleanup /AFTER
→ No improvement yet
BUT if we keep service running between tests:
Test 1: Service stopped → start(30s) → restore → cleanup /AFTER
Test 2: Service running → restore(0s saved!) → cleanup /AFTER
→ Save 30s on subsequent tests!
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Critical fix for service auto-start behavior:
Problem (identified by user):
- When Oracle service starts, it automatically tries to start instance
- Uses configured PFILE which references control files
- After cleanup, control files don't exist
- Instance ends up in partial/error state
- STARTUP NOMOUNT may fail or behave unexpectedly
Root Cause:
- Oracle service on Windows has auto-start behavior
- Service startup takes ~30s trying to start instance
- Without valid control files, instance is partially started
- This interferes with manual STARTUP NOMOUNT
Solution:
Before STARTUP NOMOUNT, explicitly clean any existing instance:
```sql
SHUTDOWN ABORT; -- Clean any partial instance
STARTUP NOMOUNT PFILE='...'; -- Fresh clean start
```
Implementation:
- Use WHENEVER SQLERROR CONTINUE (SHUTDOWN may error if no instance)
- Explicit SHUTDOWN ABORT before NOMOUNT
- Ensures clean instance state for RMAN restore
- Service running + clean NOMOUNT instance = ready for restore
User requirement met: Instance in NOMOUNT state (not mounted/open)
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Critical fix - service MUST be running for SQL*Plus connection:
Problem (confirmed by user):
- After cleanup, service is stopped
- sqlplus / as sysdba → ORA-12560: TNS:protocol adapter error
- Start-Service blocks indefinitely (user saw 25+ warnings)
- Service takes ~30 seconds to start
Previous attempt (WRONG):
- Assumed SQL*Plus works with stopped service ✗
- User proved ORA-12560 occurs when service stopped ✓
Correct Solution:
- Start service in background job (non-blocking)
- Poll service status every 3 seconds
- Timeout after 60 seconds (2x expected startup time)
- Progress logging every 15 seconds
- Cleanup background job when done
Implementation:
```powershell
Start-Job { Start-Service OracleServiceROA }
while (elapsed < 60s) {
if (service.Status == Running) → break
sleep 3s
}
```
Result:
- Service starts in ~30s (user confirmed)
- Script doesn't block
- SQL*Plus can connect successfully
- Graceful fallback if timeout exceeded
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Critical fix for service preservation:
Problem:
- After cleanup, Oracle service is stopped
- Start-Service attempts to start the instance automatically
- Without database files, service startup hangs indefinitely
- PowerShell Start-Service blocks waiting for service to start
Root Cause:
- Oracle service on Windows tries to auto-start the instance
- With no controlfile/database files, it cannot start
- Start-Service waits forever (user reported 25+ warnings)
Solution:
- Do NOT attempt to start the stopped service
- SQL*Plus can connect '/ as sysdba' even if service is stopped
- STARTUP NOMOUNT will manually start the instance
- This is the correct Oracle workflow for restore from zero
Windows SQL*Plus requirements:
✓ ORACLE_SID set (we set this)
✓ Service exists in registry (preserved after cleanup)
✓ ORACLE_HOME set (we set this)
✗ Service running NOT required for NOMOUNT startup
The service will naturally transition to Running state when
STARTUP NOMOUNT successfully starts the instance.
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Performance optimization for VM startup:
Before: Fixed 180s wait regardless of actual boot time
After: Intelligent polling with early exit when VM is ready
Implementation:
- Poll every 5 seconds (max 180s timeout)
- Check 1: VM running status in Proxmox (qm status)
- Check 2: SSH connectivity test
- Check 3: PowerShell availability (what we actually need)
- Exit immediately when all checks pass
- Progress logging every 30 seconds
- Fallback: Continue after 180s with warning
Benefits:
- Fast VM boot (30s) → saves 150s (2min 30s)
- Normal VM boot (60s) → saves 120s (2min)
- Slow VM boot → 180s (same as before)
- More robust: verifies SSH+PowerShell actually work
Average expected improvement: 60-120 seconds per test
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Critical fixes and improvements:
1. Database verification fix (robust):
- Use Select-String -Quiet to get True/False boolean
- Convert PowerShell boolean to bash-friendly format
- Check for 'READ WRITE' in entire sqlplus output
- Eliminates false negatives from text parsing issues
2. Collect FULL RMAN restore log:
- Removed -Head 200 limitation
- Now sends complete RMAN log in email
- Better debugging with full context
- Updated templates: "first 200 lines" → "complete"
3. Add bash script log to email notifications:
- Include last 100 lines of bash execution log
- Separate "RMAN Restore Log" and "Bash Script Log" sections
- Both text and HTML templates updated
- Shows script flow and any bash-level errors
This fixes the issue where 42,625 tables were restored successfully
but test reported FAILED due to query output format mismatch.
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Critical fixes for false negatives in DR test reporting:
1. Database verification fix:
- Changed from 'findstr' (CMD) to 'Select-String' (PowerShell native)
- findstr was failing in PowerShell context causing db_status to be empty
- Result: DB with 42,625 tables was incorrectly reported as FAILED
2. Restore log collection fix:
- Changed from 'type' (CMD) to 'Get-Content' (PowerShell native)
- type command doesn't work through SSH PowerShell context
- Added -ErrorAction SilentlyContinue for cleaner error handling
- Simplified fallback logic using [-z] instead of string matching
Both issues were caused by mixing CMD commands in PowerShell context.
Now uses PowerShell-native commands throughout for consistency.
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Adjust db_recovery_file_dest_size in auto-generated PFILE:
- Previous: 20G
- New: 50G
- Reason: Provide more space for RMAN restore operations and backups
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Enhancement to rman_restore_from_zero.ps1:
- Auto-generate initROA.ora if not found at service creation
- Uses exact tested configuration from initROA.ora:
- memory_target=1024M (tested DR VM allocation)
- _allow_resetlogs_corruption=TRUE (critical for DR restore!)
- control_files in oradata + recovery_area
- Standard Oracle 19c parameters for DR environment
Benefits:
- Script is now fully self-sufficient
- No manual PFILE setup required
- DR VM can be restored from completely clean state
- Uses battle-tested configuration
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Performance improvements:
- cleanup_database.ps1: Skip service deletion (saves ~25s per test)
- Remove oradim -delete, sc.exe delete, registry cleanup
- Add SPFILE deletion to ensure PFILE-based startup
- Service now persists between tests for reuse
- rman_restore_from_zero.ps1: Smart service check (saves ~15s per test)
- Check if service exists before creating
- Skip oradim -new if service already present
- Only create service on first run or if missing
Total time savings: ~40 seconds per weekly DR test
Service lifecycle: Created once, reused indefinitely until manual cleanup
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
- Fix CROSSCHECK BACKUP command to execute after database is mounted
- Correct CATALOG command to use recovery_area instead of F:\ path
- Add robust backup file validation with detailed error reporting
- Improve file-by-file backup copying with individual error tracking
- Enhance restore log collection for both success and failure scenarios
- Fix database verification to check OPEN_MODE instead of STATUS
- Add comprehensive directory and permissions error handling
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
rman_restore_from_zero.ps1:
- Add -TestMode switch parameter
- TestMode (weekly DR test): Skip service/listener config, only verify restore works
- Standalone mode: Full config with SPFILE + Listener for production use
weekly-dr-test-proxmox.sh:
- Call restore script with -TestMode flag
- Avoids service recreation and SSH disconnect during tests
Benefits:
- Weekly tests are faster and cleaner (no service restart)
- Manual restore prepares system for production use
- No more 'Broken pipe' errors during tests
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Remove service delete/recreate at step 3.3 that was causing 'Broken pipe' error
Service is already configured with auto-start at step 2.1 - no need to recreate
Issue: oradim -delete was killing running database and breaking SSH connection
Solution: Skip recreation, service already has correct auto-start configuration
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
weekly-dr-test-proxmox.sh:
- Replace Unix commands (echo, grep) with PowerShell equivalents
- Use PowerShell Select-String for database status verification
- Fix table count query to work properly through SSH
rman_restore_from_zero.ps1:
- Set Oracle service to AUTOMATIC startup (was manual)
- Set Listener service to AUTOMATIC startup
- Auto-start Listener after database restore
- Add fallback to lsnrctl if service start fails
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
