we have the following use case: Every time a certain key expires, we need to get notified and do something, based on it's value. But when redis fires the expired
event, the key was already removed from the db when we try to access it later on, which is expected of course.
Now is there a way to access the entry again, after it expired? I guess not.
So second option: Is there a way to tell redis to publish the whole value object instead of just the key when sending those events? I guess it could be added through Lua, but I'd interested in an easier option, if possible. We also need this behaviour for other events, we basically need all notifications to publish the value, not the key (we could do a GET
once the event was received, but we want to get around the second call, primarily to have an atomic process, since the value could have changed between publishing the event and doing the GET
to retrieve the value).
Hope it's understandable. Maybe we can't see the obvious, so thanks in advance!
After the timeout has expired, the key will automatically be deleted.
Hi, it is not possible, either use a different top-level key for that specific field, or store along with the filed another field with an expire time, fetch both, and let the application understand if it is still valid or not based on current time.
Redis keyspace notifications allow you to subscribe to PubSub channels. Through the channel, clients receive published events when a Redis command or data alteration occurs. These notifications are useful when an application must respond to changes that occur to the value stored in a particular key or keys.
The feature that Eli linked to allows you to listen when a key expires. However, it does not give you the value of the key. Futhermore, based on the filed github issue it does not look like you can expect to have this feature built in anytime soon if ever. The solution I use is to create a special "shadow" expiration key that is linked to the key where you have an actual value.
So lets say you have a key called testkey
and it has an integer value of 100
. Furthermore, the key will expire after 10 seconds at which point you want to get the value of the key. (Maybe you were incrementing the key during the 10 seconds it existed).
First you need to setup listening for keyspace events. In particular you want to listen for expired
events. You can do this from your config or use the config set
command in redis. (see here for more info: http://redis.io/topics/notifications)
CONFIG SET notify-keyspace-events Ex
Now you can subscribe to a special keyevent
channel where you will be notified that the key expired.
SUBSCRIBE __keyevent@0__:expired
The format of the channel to subscribe to is __keyevent@<db>__:<eventName>
. In our example we're assuming we're working with the default database 0 and we want to listen for the expired
event.
When the testkey
expires you will now get a message in the __keyevent__
channel where the message is the name of the key that expired. Of course at this point the key is gone so we can no longer access the value! The solution is to use a special expiration key.
When you create your testkey
also create a special expiring "shadow" key (don't expire the actual testkey
). For example:
SET testkey 100 SET shadowkey:testkey "" EX 10
Now in the __keyevent@0__:expired
channel you will get a message telling you that the key shadowkey:testkey
expired. Take the value of the message (which is the name of the key), split on the colon (or whatever separator you decide to use), and then manually get the value of the key and delete it.
// set your key value SET testkey 100 //set your "shadow" key, note the value here is irrelevant SET shadowkey:testkey "" EX 10 // Get an expiration message in the channel __keyevent@0__:expired // Split the key on ":", take the second part to get your original key // Then get the value and do whatever with it GET testkey // Then delete the key DEL testkey
Note that the value of the shadowkey isn't used so you want to use the smallest possible value which according to this answer (Redis store key without a value) is an empty string ""
. It's a little more work to setup but the above system does exactly what you need. The overhead is a few extra commands to actually retrieve and delete your key plus the storage cost of an empty key.
If you're on 2.8, you can try out this new feature (also referenced at this page). It's definitely unstable and doesn't seem well tested, but if you're on 2.8 anyway...
Short intro from the issue page:
An interesting feature in databases with a key-value data model (Redis does not perfectly fit this definition as the values are complex data structure, but the outer layer of Redis is definitely a key-value business) is the ability to subscribe in some way to the stream of events targeting a given key.
For instance I may be interested to see when the key foo is deleted or modified in some way, or to get the names of all the keys with an expire set (using the EXPIRE command) that Redis is evicting from the dataset because their time to live dropped to zero.
This feature was requested many times by the Redis user base, however so far we never reached a point where the proposed API (including proposals made by myself) seemed to fit well into the Redis design. This feature request will try to describe a new design that is extremely simple to use, to implement, and that fits well in the Redis story.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With