I have a lot of analytics data that I'm adding to redis. I plan on incrementally moving the data out of redis and into my database.
I know I can use KEYS [the_key]:* to get all keys that match. For example, I can do that to get the following:
127.0.0.1:6379> KEYS c_Track:*
1) "c_Track:6c93a5c1-77e9-4c4a-9232-bf182713a02e"
2) "c_Track:2c9d99c2-af37-4de9-ac64-b48f339e97a9"
3) "c_Track:9e7fd190-86d9-4b4a-9a70-7bf4c7768eef"
4) "c_Track:7f2d2e98-7440-4fd7-a80a-2af309ab15a4"
Is there a recommended way to get these values easily? I can get the keys, but how can I get all the values as well? I can loop through the keys to get the values, but is there some one-shot method for doing this?
Also I know I shouldn't use keys, but this is just an example. Thanks
Thanks
Also I know I shouldn't use keys
So don't. Use SCAN instead.
is there some one-shot method for doing this?
No, not as a core Redis command, but given the need this is fairly simple to achieve with a server-side Lua script. For example, assuming that your values are strings, you could do something like the following:
local cursor = tonumber(ARGV[1])
local pattern = ARGV[2]
local scan = redis.call('SCAN', cursor, 'MATCH', pattern)
for i, v in ipairs(scan[2]) do
local val = redis.call('GET', v)
scan[2][i] = { v, val }
end
return scan
Assuming that this script is saved under "scan.lua", you can run it as follows:
$ redis-cli SET foo bar
OK
$ redis-cli SET baz qaz
OK
$ redis-cli --eval scan.lua , 0 "*"
1) "0"
2) 1) 1) "baz"
2) "qaz"
2) 1) "foo"
2) "bar"
To scan your entire keyspace, call the script with the returned cursor until it returns 0.
Notes:
1) If your keys are of different types, you should change the script accordingly (e.g. https://github.com/itamarhaber/redis-lua-scripts/blob/master/scanfetch.lua).
2) While this script goes against the common recommendation of generating key names inside a script, it is still safe to run as SCAN returns keys that are in the server's keyspace (whether single-instance or clustered).
Related
Cannot find an easy solution. Migrate command seems not to be the solution.
Persist, delete them, create one single db and load the backup maybe?
There is the SWAPDB that allows you to swap them.
There is no out-of-the-box support for merging, since there may be conflicts. If there are two keys with the same name on both databases, it needs to be resolved.
The MOVE commands allows you to move a key from the current DB to a destination DB.
You can use a combination of SCAN and MOVE to get the merge done. I suggest SCAN instead of KEYS because if you're doing this in production, you don't want to block the server for long periods of time.
Here a Lua script to do this:
local result = redis.call('SCAN', ARGV[1])
local conflict = {}
local moved = {}
for _,key in ipairs(result[2]) do
if redis.call('MOVE', key, ARGV[2]) == 0 then
table.insert(conflict, key)
else
table.insert(moved, key)
end
end
result[2] = conflict
result[3] = moved
return result
You call this with parameters:
numkeys: 0
ARGV[1]: the cursor for SCAN, initially 0.
ARGV[2]: the destination database. E.g. 2
Make sure to SELECT the source database first.
Here a example first run:
> EVAL "local result = redis.call('SCAN', ARGV[1]) \n local conflict = {} \n local moved = {} \n for _,key in ipairs(result[2]) do \n if redis.call('MOVE', key, ARGV[2]) == 0 then \n table.insert(conflict, key) \n else \n table.insert(moved, key) \n end \n end \n result[2] = conflict \n result[3] = moved \n return result" 0 0 2
1) "56"
2) 1) "key:6"
3) 1) "key:79"
2) "key:96"
3) "key:44"
4) "key:91"
5) "key:82"
6) "key:25"
7) "key:22"
8) "key:41"
9) "key:73"
Here the SCAN found 10 keys, returned next cursor 56. key:6 was found on the destination, so it was not moved. The third entry is the list of keys actually moved.
As with SCAN, you need to call multiple times until the return cursor is 0. And remember added keys in between SCAN calls may or may not be returned.
You can modify the Lua script to:
Implement your merge logic, like delete the key on the destination and then move, and still report the deleted key on the destination separately.
Change SCAN to use a larger COUNT if you want more keys covered by call.
Replace SCAN with KEYS to do it on one go.
I'd like to periodically scan thru a redis instance for keys that changed since the last scan. in between the scans i don't want to process the keys.
eg one key could get a thousand updates between scans. i care for the most recent value only when doing the next periodic scan.
There is no built-in way in Redis to achieve that (yet).
You could, for example, recode your app and add some sort of a way to track updates. For example, wherever you're calling SET foo bar, also call ZADD updated <timestamp> foo. Then, you can use the 'updated' Sorted Set to retrieve updated keys.
Alternatively, you can try using RedisGears to automate the tracking part (for starters). Assuming that you have RedisGears running (i.e. docker run -it -p 6379:6379 redislabs/redisgears), you can do something like the following:
$ cat gear.py
def addToUpdatedZset(x):
import time
now = time.time()
execute('ZADD', 'updated', now, x['key'])
return x
GB().filter(lambda x: x['key'] != 'updated').foreach(addToUpdatedZset).register('*')
$ redis-cli RG.PYEXECUTE "$(cat gear.py)"
OK
$ redis-cli
127.0.0.1:6379> KEYS *
(empty list or set)
127.0.0.1:6379> SET foo bar
OK
127.0.0.1:6379> KEYS *
1) "updated"
2) "foo"
127.0.0.1:6379> ZRANGE updated 0 -1 WITHSCORES
1) "foo"
2) "1559339877.1392548"
127.0.0.1:6379> SET baz qux
OK
127.0.0.1:6379> KEYS *
1) "updated"
2) "baz"
3) "foo"
127.0.0.1:6379> ZRANGE updated 0 -1 WITHSCORES
1) "foo"
2) "1559339877.1392548"
3) "baz"
4) "1559339911.5493586"
My redis collection contains many keys
I want to be able to flush them all except all the keys that start with:
"configurations::"
is this possible?
You can do this
redis-cli KEYS "*" | grep -v "configurations::" | xargs redis-cli DEL
List all keys into the redis, remove from the list keys that contains "configurations::" and delete them from the redis
Edit
As #Sergio Tulentsev notice it keys is not for use in production. I used this python script to remove keys on prodution redis. I stoped replication from master to slave before call the script.
#!/usr/bin/env python
import redis
import time
pattern = "yourpattern*"
poolSlave = redis.ConnectionPool(host='yourslavehost', port=6379, db=0)
redisSlave = redis.Redis(connection_pool=poolSlave)
poolMaster = redis.ConnectionPool(host='yourmasterhost', port=6379, db=0)
redisMaster = redis.Redis(connection_pool=poolMaster)
cursor = '0'
while cursor != 0:
cursor, data = redisSlave.scan(cursor, pattern, 1000)
print "cursor: "+str(cursor)
for key in data:
redisMaster.delete(key)
print "delete key: "+key
# reduce call per second on production server
time.sleep(1)
The SCAN & DEL approach (as proposed by #khanou) is the best ad-hoc solution. Alternatively, you could keep an index of all your configurations:: key names with a Redis Set (simply SADD the key's name to it whenever you create a new configurations:: key). Once you have this set you can SSCAN it to get all the relevant key names more efficiently (don't forget to SREM from it whenever you DEL though).
Yes, it's possible. Enumerate all the keys, evaluate each one and delete if it fits the criteria for deletion.
There is no built-in redis command for this, if this is what you were asking.
It might be possible to cook up a Lua script that will do this (and it'll look to your app that it's a single command), but still it's the same approach under the hood.
In Redis, keys user* will print all keys starting with user.
For example:
keys user*
1) "user2"
2) "user1"
Now, I want all keys that don't start with user to be printed.
How could I do that?
IMPORTANT: always use SCAN instead of (the evil) KEYS
Redis' pattern matching is somewhat functionally limited (see the implementation of stringmatchlen in util.c) and does not provide that which you seek ATM. That said, consider the following possible routes:
Extend stringmatchlen to match your requirements, possibly submitting it as a PR.
Consider what you're trying to do - fetching a subset of keys is always going to be inefficient unless you index them, consider tracking the names of all non-user keys (i.e.g. in a Redis Set) instead.
If you are really insistent on scanning the entire keyspace and match against negative patterns, one way to accomplish that is with a little bit of Lua magic.
Consider the following dataset and script:
127.0.0.1:6379> dbsize
(integer) 0
127.0.0.1:6379> set user:1 1
OK
127.0.0.1:6379> set use:the:force luke
OK
127.0.0.1:6379> set non:user a
OK
Lua (save this as scanregex.lua):
local re = ARGV[1]
local nt = ARGV[2]
local cur = 0
local rep = {}
local tmp
if not re then
re = ".*"
end
repeat
tmp = redis.call("SCAN", cur, "MATCH", "*")
cur = tonumber(tmp[1])
if tmp[2] then
for k, v in pairs(tmp[2]) do
local fi = v:find(re)
if (fi and not nt) or (not fi and nt) then
rep[#rep+1] = v
end
end
end
until cur == 0
return rep
Output - first time regular matching, 2nd time the complement:
foo#bar:~$ redis-cli --eval scanregex.lua , "^user"
1) "user:1"
foo#bar:~$ redis-cli --eval scanregex.lua , "^user" 1
1) "use:the:force"
2) "non:user"
#Karthikeyan Gopall you nailed it in your comment above and this saved me a bunch of time. Thanks!
Here's how you can use it in various combinations to get what you want:
redis.domain.com:6379[1]> set "hello" "foo"
OK
redis.domain.com:6379[1]> set "hillo" "bar"
OK
redis.domain.com:6379[1]> set "user" "baz"
OK
redis.domain.com:6379[1]> set "zillo" "bash"
OK
redis.domain.com:6379[1]> scan 0
1) "0"
2) 1) "zillo"
2) "hello"
3) "user"
4) "hillo"
redis.domain.com:6379[1]> scan 0 match "[^u]*"
1) "0"
2) 1) "zillo"
2) "hello"
3) "hillo"
redis.domain.com:6379[1]> scan 0 match "[^u^z]*"
1) "0"
2) 1) "hello"
2) "hillo"
redis.domain.com:6379[1]> scan 0 match "h[^i]*"
1) "0"
2) 1) "hello"
According to redis keys documentation the command supports glob style patterns, not regular expressions.
and if you look at the documentation, you'll see that the "!" character is not special as opposites to regular expressions.
Here is a simple test I ran in my own db:
redis 127.0.0.1:6379> set a 0
OK
redis 127.0.0.1:6379> set b 1
OK
redis 127.0.0.1:6379> keys *
1) "a"
2) "b"
redis 127.0.0.1:6379> keys !a
(empty list or set) // I expected "b" here
redis 127.0.0.1:6379> keys !b
(empty list or set) // I expected "a" here
redis 127.0.0.1:6379> keys [!b]
1) "b"
redis 127.0.0.1:6379> keys [b]
1) "b"
redis 127.0.0.1:6379> keys [ab]
1) "a"
2) "b"
redis 127.0.0.1:6379> keys ![b]
(empty list or set)
So I just don't think what you are trying to achieve is possible via the keys command.
Besides, the keys command is not very suitable for production environment as it locks your whole redis database.
I would recommend getting all the keys with the scan command, store them locally, and then remove them using LUA
Here's a trick to achieve this with native redis commands (no need for Lua scripts or anything).
If you are able to control the timing of when you insert the new keys (the ones you want to keep, deleting all other stuff like in your question), you can:
Before setting the new keys, set the expiration to all existing keys (by pattern or everything) to expire right now (see how)
Load the new keys
Redis will automatically delete all the older keys and you will be left just with the new ones you want.
You also can print all keys and pass it to grep. For example:
redis-cli -a <password> keys "*" | grep -v "user"
I have a string type key value store in redis having keys like this--
/url-pattern/url-slug-1
/url-pattern/url-slug-2
/url-pattern/url-slug-3
/url-pattern/url-slug-4 ...
I can retrieve all the keys of /url-pattern/ using a wild card query like this --
keys /url-pattern/*
I would like to retrieve the values of all keys corresponding to this wildcard /url-pattern/*
I tried this
mget /url-pattern/*
1) (nil)
but it doesnt returned the array as expected.
How can I retrieve the values of all keys corresponding to /url-pattern/*
I also want to do a sum on the values, but I think there is no such thing called SUM() in redis
MGET accepts multiple arguments where each a key name. It does not do key name patterns.
What you could do is first fetch all the relevant key names (do not use KEYS, use SCAN instead) and then fetch their values with an MGET.
Here is an updated answer for 2015.
If you can upgrade Redis above 2.8, the SCAN command with MATCH will work for this. Before that version, not so much, and do NOT use the KEYS command except in a development environment.
http://redis.io/commands/scan
Example on command line:
$ redis-cli
127.0.0.1:6379> scan match V3.0:*
(error) ERR invalid cursor
127.0.0.1:6379> scan 0 match V3.0:*
1) "0"
2) 1) "V3.0:UNITTEST55660BC7E0C5B"
2) "V3.0:shop.domain.com:route"
3) "V3.0:UNITTEST55660BC4A2548"
127.0.0.1:6379> scan 0 match V1.0:*
1) "0"
2) (empty list or set)
127.0.0.1:6379> scan 0 match V3.0:*
1) "0"
2) 1) "V3.0:UNITTEST55660BC7E0C5B"
2) "V3.0:shop.domain.com:route"
3) "V3.0:UNITTEST55660BC4A2548"
Example in PHP:
// Initialize our iterator to NULL
$iterate = null;
// retry when we get no keys back
$redis->setOption(Redis::OPT_SCAN, Redis::SCAN_RETRY);
while ($arr_keys = $redis->scan($iterate, 'match:*')) {
foreach ($arr_keys as $str_key) {
echo "Here is a key: $str_key\n";
}
echo "No more keys to scan!\n";
}
Note, php code is not tested and from the core documentation for example here. Production code would need to be modified depending on the keys needed to look up.
For those on Ubuntu here are the instructions to upgrade php5-redis:
Download the 2.2.7 package here: http://pecl.php.net/package/redis
$ php -i | grep Redis
Redis Support => enabled
Redis Version => 2.2.4
Follow instructions in README to phpize, configure, make install
Create a symlink for command line cli package: cd /etc/php5/cli/conf.d && sudo ln -s ../../mods-available/redis.ini 20-redis.ini
$ php -i | grep Redis
Redis Support => enabled
Redis Version => 2.2.7
There is NO command available in REDIS which can return values from wildcard keys.
If you see the documentation for KEYS command: http://redis.io/commands/keys, it says
Consider KEYS as a command that should only be used in production
environments with extreme care. It may ruin performance when it is
executed against large databases. This command is intended for
debugging and special operations. Don't use KEYS in your regular
application code.
I don't know your business use case, but looks like you may have to use different data structure for this requirement. You can use list or set to store similar url patterns.