initial seeding
This commit is contained in:
332
list_db_snapshots.py
Normal file
332
list_db_snapshots.py
Normal file
@@ -0,0 +1,332 @@
|
||||
#!/usr/bin/env python3
|
||||
|
||||
import json
|
||||
import sys
|
||||
import os
|
||||
from datetime import datetime
|
||||
from rsc_auth import RSCAuth
|
||||
from rsc_graphql import RSCGraphQL
|
||||
from tabulate import tabulate
|
||||
|
||||
def find_database_by_name_or_id(identifier):
|
||||
"""Find database by name or ID and return its details"""
|
||||
auth = RSCAuth()
|
||||
gql = RSCGraphQL(auth)
|
||||
|
||||
# Check if identifier looks like a UUID (contains hyphens)
|
||||
if '-' in identifier:
|
||||
# It's likely a database ID
|
||||
query = """
|
||||
query OracleDatabase($fid: UUID!) {
|
||||
oracleDatabase(fid: $fid) {
|
||||
dbUniqueName
|
||||
id
|
||||
cluster {
|
||||
id
|
||||
name
|
||||
}
|
||||
logicalPath {
|
||||
fid
|
||||
name
|
||||
objectType
|
||||
}
|
||||
}
|
||||
}
|
||||
"""
|
||||
variables = {"fid": identifier}
|
||||
else:
|
||||
# It's a database name
|
||||
query = """
|
||||
query OracleDatabases($filter: [Filter!]) {
|
||||
oracleDatabases(filter: $filter) {
|
||||
nodes {
|
||||
dbUniqueName
|
||||
id
|
||||
cluster {
|
||||
id
|
||||
name
|
||||
}
|
||||
logicalPath {
|
||||
fid
|
||||
name
|
||||
objectType
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
"""
|
||||
variables = {
|
||||
"filter": [
|
||||
{"texts": [identifier], "field": "NAME_EXACT_MATCH"},
|
||||
{"texts": ["false"], "field": "IS_RELIC"},
|
||||
{"texts": ["false"], "field": "IS_REPLICATED"}
|
||||
]
|
||||
}
|
||||
|
||||
response = gql.query(query, variables)
|
||||
|
||||
if '-' in identifier:
|
||||
# Direct ID lookup
|
||||
db = response['data']['oracleDatabase']
|
||||
if not db:
|
||||
raise ValueError(f"Database with ID '{identifier}' not found")
|
||||
return db
|
||||
else:
|
||||
# Name lookup
|
||||
databases = response['data']['oracleDatabases']['nodes']
|
||||
if not databases:
|
||||
raise ValueError(f"No databases found with name '{identifier}'")
|
||||
if len(databases) > 1:
|
||||
print(f"Multiple databases found with name '{identifier}':")
|
||||
for db in databases:
|
||||
host_name = db['logicalPath'][0]['name'] if db['logicalPath'] else 'Unknown'
|
||||
print(f" - {db['dbUniqueName']} (ID: {db['id']}, Host: {host_name})")
|
||||
raise ValueError("Please specify the database ID instead")
|
||||
return databases[0]
|
||||
|
||||
def get_recoverable_ranges(db_id):
|
||||
"""Get recoverable ranges for a database"""
|
||||
auth = RSCAuth()
|
||||
gql = RSCGraphQL(auth)
|
||||
|
||||
query = """
|
||||
query OracleDatabaseRecoverableRangesQuery($fid: String!) {
|
||||
oracleRecoverableRanges(
|
||||
input: {id: $fid, shouldIncludeDbSnapshotSummaries: true}
|
||||
) {
|
||||
data {
|
||||
beginTime
|
||||
endTime
|
||||
status
|
||||
dbSnapshotSummaries {
|
||||
databaseName
|
||||
isValid
|
||||
hostOrRacName
|
||||
baseSnapshotSummary {
|
||||
id
|
||||
date
|
||||
isOnDemandSnapshot
|
||||
replicationLocationIds
|
||||
archivalLocationIds
|
||||
}
|
||||
}
|
||||
__typename
|
||||
}
|
||||
__typename
|
||||
}
|
||||
oracleMissedRecoverableRanges(input: {id: $fid}) {
|
||||
data {
|
||||
beginTime
|
||||
endTime
|
||||
__typename
|
||||
}
|
||||
__typename
|
||||
}
|
||||
}
|
||||
"""
|
||||
|
||||
variables = {"fid": db_id}
|
||||
response = gql.query(query, variables)
|
||||
|
||||
return response['data']
|
||||
|
||||
def get_cluster_name(cluster_id):
|
||||
"""Get cluster name by ID"""
|
||||
if not cluster_id:
|
||||
return "Unknown"
|
||||
|
||||
# Cache cluster names to avoid repeated queries
|
||||
if not hasattr(get_cluster_name, '_cache'):
|
||||
get_cluster_name._cache = {}
|
||||
|
||||
if cluster_id in get_cluster_name._cache:
|
||||
return get_cluster_name._cache[cluster_id]
|
||||
|
||||
try:
|
||||
auth = RSCAuth()
|
||||
gql = RSCGraphQL(auth)
|
||||
|
||||
query = """
|
||||
query ListClusters {
|
||||
allClusterConnection {
|
||||
nodes {
|
||||
id
|
||||
name
|
||||
}
|
||||
}
|
||||
}
|
||||
"""
|
||||
|
||||
response = gql.query(query)
|
||||
clusters = response['data']['allClusterConnection']['nodes']
|
||||
|
||||
# Find the cluster with matching ID
|
||||
for cluster in clusters:
|
||||
if cluster['id'] == cluster_id:
|
||||
name = cluster['name']
|
||||
get_cluster_name._cache[cluster_id] = name
|
||||
return name
|
||||
|
||||
# If not found, return fallback
|
||||
return f"Cluster-{cluster_id[:8]}"
|
||||
except:
|
||||
return f"Cluster-{cluster_id[:8]}"
|
||||
|
||||
def get_location_name(location_type, location_ids):
|
||||
"""Get human-readable location names"""
|
||||
if not location_ids:
|
||||
return []
|
||||
|
||||
names = []
|
||||
for loc_id in location_ids:
|
||||
if location_type == "replication":
|
||||
name = get_cluster_name(loc_id)
|
||||
names.append(f"Replicated: {name}")
|
||||
elif location_type == "archival":
|
||||
# For archival, we might not have cluster info, so use a generic name
|
||||
names.append(f"Archive: {loc_id[:8]}...")
|
||||
else:
|
||||
names.append(f"{location_type}: {loc_id[:8]}...")
|
||||
|
||||
return names
|
||||
|
||||
def format_timestamp(timestamp):
|
||||
"""Format ISO timestamp to readable format"""
|
||||
try:
|
||||
dt = datetime.fromisoformat(timestamp.replace('Z', '+00:00'))
|
||||
return dt.strftime('%Y-%m-%d %H:%M:%S UTC')
|
||||
except:
|
||||
return timestamp
|
||||
|
||||
def list_snapshots_and_ranges(identifier):
|
||||
"""List snapshots and recovery ranges for a database"""
|
||||
try:
|
||||
# Find the database
|
||||
db = find_database_by_name_or_id(identifier)
|
||||
|
||||
print(f"Database: {db['dbUniqueName']}")
|
||||
print(f"ID: {db['id']}")
|
||||
cluster_name = db['cluster']['name'] if db['cluster'] else 'Unknown'
|
||||
print(f"Cluster: {cluster_name}")
|
||||
host_name = db['logicalPath'][0]['name'] if db['logicalPath'] else 'Unknown'
|
||||
print(f"Host: {host_name}")
|
||||
print("-" * 80)
|
||||
|
||||
# Get recoverable ranges
|
||||
ranges_data = get_recoverable_ranges(db['id'])
|
||||
|
||||
# Display recoverable ranges
|
||||
recoverable_ranges = ranges_data.get('oracleRecoverableRanges', {}).get('data', [])
|
||||
if recoverable_ranges:
|
||||
print(f"\nRecoverable Ranges ({len(recoverable_ranges)} found):")
|
||||
table_data = []
|
||||
for range_item in recoverable_ranges:
|
||||
table_data.append([
|
||||
format_timestamp(range_item['beginTime']),
|
||||
format_timestamp(range_item['endTime']),
|
||||
range_item.get('status', 'Unknown')
|
||||
])
|
||||
|
||||
headers = ['Begin Time', 'End Time', 'Status']
|
||||
print(tabulate(table_data, headers=headers, tablefmt='grid'))
|
||||
|
||||
# Display snapshots by location
|
||||
print(f"\nSnapshot Distribution:")
|
||||
location_snapshots = {}
|
||||
|
||||
for range_item in recoverable_ranges:
|
||||
snapshots = range_item.get('dbSnapshotSummaries', [])
|
||||
for snapshot in snapshots:
|
||||
base = snapshot.get('baseSnapshotSummary', {})
|
||||
|
||||
# Group by replication locations
|
||||
repl_locations = base.get('replicationLocationIds', [])
|
||||
arch_locations = base.get('archivalLocationIds', [])
|
||||
|
||||
# Add to local (assuming no replication/archival IDs means local)
|
||||
if not repl_locations and not arch_locations:
|
||||
loc_name = "Local (Source)"
|
||||
if loc_name not in location_snapshots:
|
||||
location_snapshots[loc_name] = []
|
||||
location_snapshots[loc_name].append({
|
||||
'id': base.get('id'),
|
||||
'date': base.get('date'),
|
||||
'isOnDemand': base.get('isOnDemandSnapshot', False),
|
||||
'host': snapshot.get('hostOrRacName', 'Unknown')
|
||||
})
|
||||
else:
|
||||
# Handle replication locations
|
||||
for repl_id in repl_locations:
|
||||
cluster_name = get_cluster_name(repl_id)
|
||||
loc_name = f"Replicated: {cluster_name}"
|
||||
if loc_name not in location_snapshots:
|
||||
location_snapshots[loc_name] = []
|
||||
location_snapshots[loc_name].append({
|
||||
'id': base.get('id'),
|
||||
'date': base.get('date'),
|
||||
'isOnDemand': base.get('isOnDemandSnapshot', False),
|
||||
'host': snapshot.get('hostOrRacName', 'Unknown')
|
||||
})
|
||||
|
||||
# Handle archival locations
|
||||
for arch_id in arch_locations:
|
||||
loc_name = f"Archived: {arch_id[:8]}..."
|
||||
if loc_name not in location_snapshots:
|
||||
location_snapshots[loc_name] = []
|
||||
location_snapshots[loc_name].append({
|
||||
'id': base.get('id'),
|
||||
'date': base.get('date'),
|
||||
'isOnDemand': base.get('isOnDemandSnapshot', False),
|
||||
'host': snapshot.get('hostOrRacName', 'Unknown')
|
||||
})
|
||||
|
||||
# Display snapshots by location
|
||||
for location, snapshots in location_snapshots.items():
|
||||
print(f"\n📍 {location} ({len(snapshots)} snapshots):")
|
||||
table_data = []
|
||||
for snap in sorted(snapshots, key=lambda x: x['date']):
|
||||
table_data.append([
|
||||
snap['id'][:8] + '...', # Truncate ID for readability
|
||||
format_timestamp(snap['date']),
|
||||
'On-Demand' if snap['isOnDemand'] else 'Policy',
|
||||
snap['host']
|
||||
])
|
||||
|
||||
headers = ['Snapshot ID', 'Date', 'Type', 'Host']
|
||||
print(tabulate(table_data, headers=headers, tablefmt='grid'))
|
||||
else:
|
||||
print("\nNo recoverable ranges found.")
|
||||
|
||||
# Display missed recoverable ranges
|
||||
missed_ranges = ranges_data.get('oracleMissedRecoverableRanges', {}).get('data', [])
|
||||
if missed_ranges:
|
||||
print(f"\nMissed Recoverable Ranges ({len(missed_ranges)} found):")
|
||||
table_data = []
|
||||
for range_item in missed_ranges:
|
||||
table_data.append([
|
||||
format_timestamp(range_item['beginTime']),
|
||||
format_timestamp(range_item['endTime'])
|
||||
])
|
||||
|
||||
headers = ['Begin Time', 'End Time']
|
||||
print(tabulate(table_data, headers=headers, tablefmt='grid'))
|
||||
else:
|
||||
print("\nNo missed recoverable ranges found.")
|
||||
|
||||
except Exception as e:
|
||||
print(f"Error: {e}", file=sys.stderr)
|
||||
sys.exit(1)
|
||||
|
||||
def main():
|
||||
if len(sys.argv) != 2:
|
||||
print("Usage: python list_db_snapshots.py <database_name_or_id>")
|
||||
print("Examples:")
|
||||
print(" python list_db_snapshots.py SCLONE")
|
||||
print(" python list_db_snapshots.py 2cb7e201-9da0-53f2-8c69-8fc21f82e0d2")
|
||||
sys.exit(1)
|
||||
|
||||
identifier = sys.argv[1]
|
||||
list_snapshots_and_ranges(identifier)
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
Reference in New Issue
Block a user