rediff - the redis diff script

rediff.py

import urlparse
import argparse
import redis
import sys
from multiprocessing import Pool
import signal

def parse_redis_url(s):
    url = urlparse.urlparse(s)
    if not url.scheme:
        s = 'redis://' + s
        url = urlparse.urlparse(s)
    if url.scheme != 'redis':
        print 'Invalid scheme %s for %s'%(url.scheme,s)
        return None
    try:
        db = int(url.path)
    except ValueError:
        db = 0        
    return (url.hostname, url.port or 6379, url.password, db)
    
def compare_string(src, dst, key):
    if src.get(key) != dst.get(key):
        return False
    return True        

def compare_hash(src, dst, key):
    h1 = src.hgetall(key)
    h2 = dst.hgetall(key)
    if set(h1) != set(h2):
        return False
        
    for k,v in h1.iteritems():
        if h2[k] != v:
            return False
    return True            

def compare_list(src, dst, key):
    if src.lrange(key, 0, -1) != dst.lrange(key, 0, -1):
        return False
    return True            

def compare_set(src, dst, key):
    if src.smembers(key) != dst.smembers(key):
        return False
    return True            

def compare_zset(src, dst, key):
    if src.zrange(key, 0, -1, withscores = True) != dst.zrange(key, 0, -1, withscores = True):
        return False
    return True
    
def compare(key):
    try:
        res = cmp_funcs[src.type(key)](src, dst, key)
        if not res:
            print "key '%s' differs"%key
        return res
    except redis.ResponseError as e:
        print "Error '%s' when comparing key: '%s', skipping"%(e, key)
        return True

def init_worker():
    signal.signal(signal.SIGINT, signal.SIG_IGN)
    
if __name__ == '__main__':
    argparser = argparse.ArgumentParser(description='Redis diff tool.')
    argparser.add_argument('--src', required='True')
    argparser.add_argument('--dst', required='True')
    
    args = argparser.parse_args()
    
    host,port,password,db = parse_redis_url(args.src)
    src = redis.Redis(host = host, port = port, password = password, db = db)
    host,port,password,db = parse_redis_url(args.dst)
    dst = redis.Redis(host = host, port = port, password = password, db = db)
    
    src_key_count = src.dbsize()
    dst_key_count = dst.dbsize()
    if dst_key_count != src_key_count:
        print "Source and destination key counts differ (%d : %d)"%(src_key_count, dst_key_count)
        
    cmp_funcs = {}
    for t in ['string', 'hash', 'list', 'set', 'zset']:
        cmp_funcs[t] = globals().get("compare_" + t)
        
    pool = Pool(50, init_worker)
    
    it = 0
    scanned = 0
    diffs = 0
    try:
        while True:
            it, keys = src.execute_command('scan', int(it), 'count', 1000)
            if int(it) == 0:
                break

            # Use multiprocess pool to compare lots of keys simultaneously
            res = pool.map(compare, keys)
            diffs += len([x for x in res if x == False])

# Straight forward non-parallel approach, slower
#            for key in keys:
#                try:
#                    if not compare(key):
#                        print "key '%s' differs"%key
#                        diffs += 1
#                except redis.ResponseError as e:
#                    print "Error '%s' when comparing key: '%s'"%(e, key)
                    
            scanned += len(keys)
            sys.stdout.write("\rprogress: {:3.2f}".format(100.0*float(scanned)/src_key_count))
            sys.stdout.flush()
    
        print 'Done. Found %d diffs'%diffs                

    except KeyboardInterrupt:
        print "Caught KeyboardInterrupt, terminating workers"
        pool.terminate()
        pool.join()        

Can the DUMP command be used to compare the values?

Can the DUMP command be used to compare the values?

Not really because values might be differently encoded internally for various reasons.

Are we able to retrieve the data that is different?

@asmaasfak the script outputs the names of the different keys. You can then query them individually. If you want you can modify the script to print the values of the differing keys.

nice work, you saved me a lot of headaches

Can the DUMP command be used to compare the values?

In general it can be, but it has several drawbacks. For example it can't be used for big collections, because it will lock your redis thread. Also it may probably have different binary formats across different major redis versions(but I am bot sure)

Not really because values might be differently encoded internally for various reasons.

We have compared about 1TB of data between redis cluster 6.0 installed on intel machines and Elasticache(engine: redis 6.2) installed on arm machines and found only 4 cases where comparing through dumps approach showed us false negative. So I am pretty sure that inside one major version of redis it can be compared pretty well.

Will this work for caches behind vpc?