sstablerepairedset

Marks SSTables as repaired or unrepaired, controlling their participation in incremental repair.

---

Synopsis

sstablerepairedset --really-set [--is-repaired | --is-unrepaired] <sstable_files>
sstablerepairedset --really-set [--is-repaired | --is-unrepaired] -f <sstable_list_file>

---

Description

sstablerepairedset modifies the repair status metadata of SSTable files. This status determines how SSTables participate in Cassandra's incremental repair mechanism:

• Repaired - SSTable has been verified consistent across replicas
• Unrepaired - SSTable needs to be included in future repair operations

This tool is essential for:

• Migrating to incremental repair - Mark existing SSTables appropriately
• Fixing repair metadata - Correct incorrect repair status
• Recovery scenarios - Reset repair state after data issues
• Anti-entropy management - Control what gets repaired

Cassandra Must Be Stopped

Cassandra must be completely stopped before running sstablerepairedset. Modifying repair status while Cassandra is active can cause data inconsistency.

---

How Repair Status Works

Repair Status Values

Status	repaired_at Value	Meaning
Unrepaired	0	Needs repair verification
Repaired	Timestamp	Verified at that time
Pending	UUID	Part of ongoing repair

---

Arguments

Argument	Description
sstable_files	One or more paths to SSTable Data.db files

---

Options

Option	Description
--really-set	Required safety flag to confirm modification
--is-repaired	Mark SSTables as repaired
--is-unrepaired	Mark SSTables as unrepaired
-f <file>	Read list of SSTable paths from file (one path per line)

The --really-set flag is mandatory to prevent accidental modification.

Repaired Timestamp Behavior

When using --is-repaired, the tool sets repaired_at to the Data.db file's last-modified time, not to the current time or an actual repair timestamp. This is important when interpreting the metadata later.

---

Examples

Mark as Repaired

# Stop Cassandra first
sudo systemctl stop cassandra

# Mark specific SSTable as repaired
sstablerepairedset --really-set --is-repaired \
    /var/lib/cassandra/data/my_keyspace/my_table-abc123/nb-1-big-Data.db

# Start Cassandra
sudo systemctl start cassandra

Mark as Unrepaired

# Mark SSTable as unrepaired
sstablerepairedset --really-set --is-unrepaired \
    /var/lib/cassandra/data/my_keyspace/my_table-abc123/nb-1-big-Data.db

Mark Multiple SSTables

# Mark all SSTables for a table as unrepaired
sstablerepairedset --really-set --is-unrepaired \
    /var/lib/cassandra/data/my_keyspace/my_table-*/*-Data.db

Using with find

# Mark all SSTables in keyspace as unrepaired
find /var/lib/cassandra/data/my_keyspace/ -name "*-Data.db" -print0 | \
    xargs -0 sstablerepairedset --really-set --is-unrepaired

Using File Input

# Create a list of SSTable paths
find /var/lib/cassandra/data/my_keyspace/ -name "*-Data.db" > /tmp/sstables.txt

# Mark all listed SSTables as unrepaired
sstablerepairedset --really-set --is-unrepaired -f /tmp/sstables.txt

Mark All SSTables on Node

#!/bin/bash
# mark_all_unrepaired.sh - Reset all repair status

DATA_DIR="/var/lib/cassandra/data"

# Find all user keyspaces (exclude system keyspaces)
for ks_dir in ${DATA_DIR}/*/; do
    ks_name=$(basename "$ks_dir")

    # Skip system keyspaces
    if [[ "$ks_name" == system* ]]; then
        continue
    fi

    echo "Processing keyspace: $ks_name"

    find "$ks_dir" -name "*-Data.db" -print0 | \
        xargs -0 -r sstablerepairedset --really-set --is-unrepaired
done

echo "All user SSTables marked as unrepaired"

---

When to Use sstablerepairedset

Scenario 1: Migrating to Incremental Repair

migrate_to_incremental.sh

#!/bin/bash
KEYSPACE="$1"

# 1. Stop Cassandra
sudo systemctl stop cassandra

# 2. Mark all existing SSTables as unrepaired
# This ensures they will be included in the first incremental repair
find /var/lib/cassandra/data/${KEYSPACE}/ -name "*-Data.db" -print0 | \
    xargs -0 sstablerepairedset --really-set --is-unrepaired

# 3. Start Cassandra
sudo systemctl start cassandra

# 4. Run incremental repair
nodetool repair -pr "$KEYSPACE"

Scenario 2: Fixing Incorrect Repair Status

# If SSTables were incorrectly marked as repaired

sudo systemctl stop cassandra

# Reset to unrepaired
sstablerepairedset --really-set --is-unrepaired \
    /var/lib/cassandra/data/my_keyspace/my_table-*/*-Data.db

sudo systemctl start cassandra

# Re-run repair
nodetool repair my_keyspace my_table

Scenario 3: After sstablescrub

# Scrub marks SSTables as unrepaired, verify status

# Check current status
sstablemetadata /var/lib/cassandra/data/my_keyspace/my_table-*/*-Data.db | \
    grep "Repaired"

# If needed, ensure consistent state
sudo systemctl stop cassandra
sstablerepairedset --really-set --is-unrepaired \
    /var/lib/cassandra/data/my_keyspace/my_table-*/*-Data.db
sudo systemctl start cassandra

# Repair to restore consistency
nodetool repair my_keyspace my_table

Scenario 4: After Data Recovery

# After restoring from backup, data needs repair verification

sudo systemctl stop cassandra

# Mark restored SSTables as unrepaired
sstablerepairedset --really-set --is-unrepaired \
    /var/lib/cassandra/data/restored_keyspace/*/*.db

sudo systemctl start cassandra

# Full repair to sync with cluster
nodetool repair restored_keyspace

Scenario 5: Switching from Full to Incremental Repair

setup_incremental_repair.sh

#!/bin/bash
# 1. Ensure all repairs are complete
nodetool repair -full

# 2. Stop Cassandra
sudo systemctl stop cassandra

# 3. Mark all as repaired (since full repair just completed)
find /var/lib/cassandra/data/ -name "*-Data.db" \
    ! -path "*/system*" -print0 | \
    xargs -0 sstablerepairedset --really-set --is-repaired

# 4. Start Cassandra
sudo systemctl start cassandra

# 5. Future repairs will be incremental
# Only new unrepaired SSTables will be included

---

Verification

Check Current Repair Status

# View repair status for all SSTables
sstablemetadata /var/lib/cassandra/data/my_keyspace/my_table-*/*-Data.db | \
    grep -E "SSTable:|Repaired"

Sample Output

SSTable: nb-1-big-Data.db
Repaired At: 1705401600000 (2024-01-16T12:00:00.000Z)

SSTable: nb-2-big-Data.db
Repaired At: 0 (unrepaired)

SSTable: nb-3-big-Data.db
Repaired At: 1705488000000 (2024-01-17T12:00:00.000Z)

Audit Script

repair_status_audit.sh

#!/bin/bash
KEYSPACE="$1"

echo "Repair Status Audit for ${KEYSPACE}"
echo "===================================="

repaired=0
unrepaired=0

for sstable in /var/lib/cassandra/data/${KEYSPACE}/*/*-Data.db; do
    status=$(sstablemetadata "$sstable" 2>/dev/null | grep "Repaired At:" | awk '{print $3}')

    if [ "$status" = "0" ]; then
        unrepaired=$((unrepaired + 1))
        echo "UNREPAIRED: $(basename $sstable)"
    else
        repaired=$((repaired + 1))
    fi
done

echo ""
echo "Summary:"
echo "  Repaired: $repaired"
echo "  Unrepaired: $unrepaired"
echo "  Total: $((repaired + unrepaired))"

---

Impact of Repair Status

Marking as Repaired

Aspect	Effect
Incremental repair	Excluded from future repairs
Compaction	Compacted only with other repaired SSTables
Anti-entropy	Considered consistent
Risk	If data actually inconsistent, won't be fixed

Marking as Unrepaired

Aspect	Effect
Incremental repair	Included in next repair
Compaction	Compacted only with other unrepaired SSTables
Anti-entropy	Will be verified against replicas
Risk	Increased repair work, but ensures consistency

Compaction Segregation

---

Troubleshooting

Permission Denied

# Run as cassandra user
sudo -u cassandra sstablerepairedset --really-set --is-unrepaired \
    /var/lib/cassandra/data/.../*-Data.db

# Or fix ownership after
sudo chown -R cassandra:cassandra /var/lib/cassandra/data/

Cassandra Still Running

# Must stop Cassandra first!
nodetool drain
sudo systemctl stop cassandra

# Verify stopped
pgrep -f CassandraDaemon  # Should return nothing

# Now safe to run
sstablerepairedset --really-set --is-unrepaired /path/to/sstable-Data.db

Missing --really-set Flag

# Error: Must provide --really-set flag

# The flag is required as a safety measure
sstablerepairedset --really-set --is-unrepaired /path/to/sstable-Data.db

Forgot --is-repaired or --is-unrepaired

# Error: Must specify either --is-repaired or --is-unrepaired

# Choose one:
sstablerepairedset --really-set --is-repaired /path/to/sstable-Data.db
# or
sstablerepairedset --really-set --is-unrepaired /path/to/sstable-Data.db

---

Best Practices

sstablerepairedset Guidelines

1. Understand implications - Wrong status causes data issues
2. Prefer unrepaired when uncertain - Safer to re-repair
3. Verify after setting - Check with sstablemetadata
4. Document changes - Track what was modified and why
5. Run repair after - Especially after marking unrepaired
6. Backup first - Snapshot before bulk changes
7. Consistent approach - Apply to all nodes in cluster

Critical Warnings

• Never mark as repaired without actual repair - Causes silent data loss
• Repair status affects compaction - Wrong status causes issues
• Cluster-wide impact - Inconsistent status causes problems
• Cannot undo easily - Wrong status requires re-repair

---

Related Commands

Command	Relationship
sstablemetadata	Verify repair status
nodetool repair	Run repair operations
nodetool repair_admin	Check repair progress
sstablescrub	Also affects repair status
sstableupgrade	Also affects repair status