Managing Solr

Solr Management Skill

Monitor, analyze, and optimize Apache Solr instances and SolrCloud clusters safely.

MANDATORY: Discovery-First Pattern

Always check cluster status and list collections before any query or config operations. Never assume collection or field names.

Phase 1: Discovery

#!/bin/bash

SOLR_URL="${SOLR_URL:-http://localhost:8983/solr}"

echo "=== System Info ==="
curl -s "$SOLR_URL/admin/info/system?wt=json" | python3 -c "
import sys, json
d = json.load(sys.stdin)
print(f\"Solr Version: {d.get('lucene',{}).get('solr-spec-version','?')}\")
print(f\"JVM: {d.get('jvm',{}).get('version','?')}\")
print(f\"Uptime: {d.get('jvm',{}).get('jmx',{}).get('upTimeMS',0)//86400000}d\")
" 2>/dev/null

echo ""
echo "=== Collections ==="
curl -s "$SOLR_URL/admin/collections?action=LIST&wt=json" | python3 -c "
import sys, json
data = json.load(sys.stdin)
for c in data.get('collections', []):
    print(f\"  {c}\")
" 2>/dev/null

echo ""
echo "=== Cluster Status ==="
curl -s "$SOLR_URL/admin/collections?action=CLUSTERSTATUS&wt=json" | python3 -c "
import sys, json
data = json.load(sys.stdin)
cluster = data.get('cluster', {})
live = cluster.get('live_nodes', [])
print(f\"Live nodes: {len(live)}\")
for name, coll in cluster.get('collections', {}).items():
    shards = coll.get('shards', {})
    print(f\"Collection: {name} | Shards: {len(shards)} | RF: {coll.get('replicationFactor','?')}\")
" 2>/dev/null

echo ""
echo "=== Core Status ==="
curl -s "$SOLR_URL/admin/cores?action=STATUS&wt=json" | python3 -c "
import sys, json
data = json.load(sys.stdin)
for name, core in data.get('status', {}).items():
    idx = core.get('index', {})
    print(f\"Core: {name} | Docs: {idx.get('numDocs',0)} | Size: {idx.get('sizeInBytes',0)//1048576}MB\")
" 2>/dev/null

Phase 1 outputs: Solr version, collection list, cluster topology, core doc counts and sizes.

Phase 2: Analysis

#!/bin/bash

SOLR_URL="${SOLR_URL:-http://localhost:8983/solr}"
COLLECTION="${1:-my_collection}"

echo "=== Schema Fields for $COLLECTION ==="
curl -s "$SOLR_URL/$COLLECTION/schema/fields?wt=json" | python3 -c "
import sys, json
for f in json.load(sys.stdin).get('fields', []):
    print(f\"  {f['name']}: {f.get('type','?')} indexed={f.get('indexed','?')} stored={f.get('stored','?')}\")
" 2>/dev/null | head -20

echo ""
echo "=== Cache Stats ==="
curl -s "$SOLR_URL/$COLLECTION/admin/mbeans?cat=CACHE&stats=true&wt=json" | python3 -c "
import sys, json
data = json.load(sys.stdin)
beans = data.get('solr-mbeans', [])
i = 0
while i < len(beans)-1:
    if isinstance(beans[i+1], dict):
        for name, info in beans[i+1].items():
            stats = info.get('stats', {})
            if stats:
                print(f\"{name}: hits={stats.get('hits','?')} lookups={stats.get('lookups','?')} hitratio={stats.get('hitratio','?')}\")
    i += 2
" 2>/dev/null

echo ""
echo "=== Query Handler Stats ==="
curl -s "$SOLR_URL/$COLLECTION/admin/mbeans?cat=QUERYHANDLER&stats=true&wt=json" | python3 -c "
import sys, json
data = json.load(sys.stdin)
beans = data.get('solr-mbeans', [])
i = 0
while i < len(beans)-1:
    if isinstance(beans[i+1], dict):
        for name, info in beans[i+1].items():
            stats = info.get('stats', {})
            if stats and stats.get('requests',0) > 0:
                print(f\"{name}: requests={stats.get('requests')} avgTime={stats.get('avgTimePerRequest','?')}ms errors={stats.get('errors',0)}\")
    i += 2
" 2>/dev/null | head -10

Output Format

SOLR ANALYSIS
=============
Version: [version] | Mode: [standalone/SolrCloud]
Collections: [count] | Live Nodes: [count]

ISSUES FOUND:
- [issue with affected collection/core]

RECOMMENDATIONS:
- [actionable recommendation]

Anti-Hallucination Rules

1. NEVER assume resource names — always discover via CLI/API in Phase 1 before referencing in Phase 2.
2. NEVER fabricate metric names or dimensions — verify against the service documentation or --help output.
3. NEVER mix CLI commands between service versions — confirm which version/API you are targeting.
4. ALWAYS use the discovery → verify → analyze chain — every resource referenced must have been discovered first.
5. ALWAYS handle empty results gracefully — an empty response is valid data, not an error to retry.

Counter-Rationalizations