I am currently developing a system that makes heavy use of redis for a series of web services.
One of the key criteria of this system is fast responses.
At present the layout (ignoring load balancers etc) is as follows:
2 x Front End Play Framework 2.x Servers
2 x Job Handling/Persistence Play Framework 2.x Servers
1 x MySQL Server
2 x Redis Servers, 1 master, 1 slave
In this setup, redis serves 2 tasks - as a shared cache and also as a message bus.
Currently the front end servers host a service which interacts in its entirety with Redis.
The front end servers try to balance reads across the pool of read servers (currently the master and 1 slave), but being Redis they need to make their writes to the master server. They handle cache updates etc by sending messages on the queues, which are picked up by the job handling servers.
The job handling servers do blocking listens (BLPOP) to the Redis write server and process tasks when necessary. They have the only connection to MySQL.
At present the read replica server is a dedicated server - more there to be able to switch it to write master if the current master fails.
I was thinking of putting a read replica slave of redis on each of the front end servers which means that read latency would be even less, and writes (messages for queues) get pushed to the write server on a separate connection.
If I need to scale, I could just add more front end servers with read slaves.
It sounds like a win/win to me as even if the write server temporarily drops out, the front end servers can still read data at least from their local slave and act accordingly.
Can anyone think of reasons why this might not be such a good idea?
I understand the advantages of this approach... but consider this: what happens when you need to scale just one component (i.e. FE server or Redis) but not the other? For example, more traffic could mean you'll need more app servers to handle it while the Redises will be considerably less loaded. On the other hand, if your dataset grows and/or more load is put on the Redises - you'll need to scale these and not the app.
The design should fit your requirements, and the simplicity of your suggested setup has a definite appeal (i.e. to scale, just add another identical lego block) but from my meager experience - anything that sounds too good to be true usually is. In the longer run, even if this works for you now, you may find yourself in a jam down the road. My advice - separate your Redis(es) from you app servers, deal with and/or work around the network and make sure each layer is available and scalable on its own right.
Related
I've found different zookeeper definitions across multiple resources. Maybe some of them are taken out of context, but look at them pls:
A canonical example of Zookeeper usage is distributed-memory computation...
ZooKeeper is an open source Apache™ project that provides a centralized infrastructure and services that enable synchronization across a cluster.
Apache ZooKeeper is an open source file application program interface (API) that allows distributed processes in large systems to synchronize with each other so that all clients making requests receive consistent data.
I've worked with Redis and Hazelcast, that would be easier for me to understand Zookeeper by comparing it with them.
Could you please compare Zookeeper with in-memory-data-grids and Redis?
If distributed-memory computation, how does zookeeper differ from in-memory-data-grids?
If synchronization across cluster, than how does it differs from all other in-memory storages? The same in-memory-data-grids also provide cluster-wide locks. Redis also has some kind of transactions.
If it's only about in-memory consistent data, than there are other alternatives. Imdg allow you to achieve the same, don't they?
https://zookeeper.apache.org/doc/current/zookeeperOver.html
By default, Zookeeper replicates all your data to every node and lets clients watch the data for changes. Changes are sent very quickly (within a bounded amount of time) to clients. You can also create "ephemeral nodes", which are deleted within a specified time if a client disconnects. ZooKeeper is highly optimized for reads, while writes are very slow (since they generally are sent to every client as soon as the write takes place). Finally, the maximum size of a "file" (znode) in Zookeeper is 1MB, but typically they'll be single strings.
Taken together, this means that zookeeper is not meant to store for much data, and definitely not a cache. Instead, it's for managing heartbeats/knowing what servers are online, storing/updating configuration, and possibly message passing (though if you have large #s of messages or high throughput demands, something like RabbitMQ will be much better for this task).
Basically, ZooKeeper (and Curator, which is built on it) helps in handling the mechanics of clustering -- heartbeats, distributing updates/configuration, distributed locks, etc.
It's not really comparable to Redis, but for the specific questions...
It doesn't support any computation and for most data sets, won't be able to store the data with any performance.
It's replicated to all nodes in the cluster (there's nothing like Redis clustering where the data can be distributed). All messages are processed atomically in full and are sequenced, so there's no real transactions. It can be USED to implement cluster-wide locks for your services (it's very good at that in fact), and tehre are a lot of locking primitives on the znodes themselves to control which nodes access them.
Sure, but ZooKeeper fills a niche. It's a tool for making a distributed applications play nice with multiple instances, not for storing/sharing large amounts of data. Compared to using an IMDG for this purpose, Zookeeper will be faster, manages heartbeats and synchronization in a predictable way (with a lot of APIs for making this part easy), and has a "push" paradigm instead of "pull" so nodes are notified very quickly of changes.
The quotation from the linked question...
A canonical example of Zookeeper usage is distributed-memory computation
... is, IMO, a bit misleading. You would use it to orchestrate the computation, not provide the data. For example, let's say you had to process rows 1-100 of a table. You might put 10 ZK nodes up, with names like "1-10", "11-20", "21-30", etc. Client applications would be notified of this change automatically by ZK, and the first one would grab "1-10" and set an ephemeral node clients/192.168.77.66/processing/rows_1_10
The next application would see this and go for the next group to process. The actual data to compute would be stored elsewhere (ie Redis, SQL database, etc). If the node failed partway through the computation, another node could see this (after 30-60 seconds) and pick up the job again.
I'd say the canonical example of ZooKeeper is leader election, though. Let's say you have 3 nodes -- one is master and the other 2 are slaves. If the master goes down, a slave node must become the new leader. This type of thing is perfect for ZK.
Consistency Guarantees
ZooKeeper is a high performance, scalable service. Both reads and write operations are designed to be fast, though reads are faster than writes. The reason for this is that in the case of reads, ZooKeeper can serve older data, which in turn is due to ZooKeeper's consistency guarantees:
Sequential Consistency
Updates from a client will be applied in the order that they were sent.
Atomicity
Updates either succeed or fail -- there are no partial results.
Single System Image
A client will see the same view of the service regardless of the server that it connects to.
Reliability
Once an update has been applied, it will persist from that time forward until a client overwrites the update. This guarantee has two corollaries:
If a client gets a successful return code, the update will have been applied. On some failures (communication errors, timeouts, etc) the client will not know if the update has applied or not. We take steps to minimize the failures, but the only guarantee is only present with successful return codes. (This is called the monotonicity condition in Paxos.)
Any updates that are seen by the client, through a read request or successful update, will never be rolled back when recovering from server failures.
Timeliness
The clients view of the system is guaranteed to be up-to-date within a certain time bound. (On the order of tens of seconds.) Either system changes will be seen by a client within this bound, or the client will detect a service outage.
Ok, so what I have are 2 web servers running inside of a Windows NLB clustered environment. The servers are identical in every respect, and as you'd expect in an NLB clustered environment, everybody is hitting the cluster name and not the individual members. We also have affinity turned off on the members in the cluster.
But, what I'm trying to do is to turn on some caching for a few large files (MP3s). It's easy enough to dial up a Redis node on one particular member and hit it, everything works like you'd expect. I can pull the data from the cache and serve it up as needed.
Now, let's add the overhead of the NLB. With an NLB in play, you may not be hitting the same web server each time. You might make your first hit to member 01, and the second hit to 02. So, I'd need a way to sync between the two servers. That way it doesn't matter which cluster member you hit, you are going to get the same data.
I don't need to worry about one cache being out of date, the only thing I'm storing in there is read only data from an internal web service.
I've only got 2 servers and it looks like redis clusters need 3. So I guess that's out.
Is this the best approach? Or perhaps there is something else better?
Reasons for redis: We only want the cache to use in-memory only. No writes to the database. Thought this would be a good fit, but need to make sure the data is available in both servers.
It's not possible to have redis multi master (writing on both). And I might say it's replication is blazing fast (check the slaveof command of Redis).
But why you need it in the same server? Access it as a service. So every node will access the actual data. If the main server goes down, the slave will promptly turn itself into a master.
One observation: you might notice that Redis makes use of disk in an async way. An append only file that it does checkpoint depending on the size from time to time so.
I have integrated twemproxy into web layer and I have 6 Elasticache(1 master , 5 read replicas) I am getting issue that the all replicas have same keys everything is same but cache hits on one replica is way more than others and I performed several load testing still on every test I am getting same result. I have separate data engine that writes on the master of this cluster and remaining 5 replicas get sync with it. So I am using twemproxy only for reading data from Elasticache not for sharding purpose. So my simple question is why i am getting 90% of hits on single read replicas of Elasticache it should distribute the hits evenly among all read replicas? right?
Thank you in advance
Twemproxy hashes everything as I recall. This means it will try to split keys among the masters you give it. If you have one master this means it hashes everything to one server. Thus, as far as it is concerned you have one server for acceptable queries. As such, it isn't helping you in this case.
If you want to have a single endpoint to distribute reads across a bank of identical slaves, you will need to put a TCP load balancer in front of the slaves and have your application talk to the IP:port of the load balancer. Common options are Nginx and HAProxy for software based ones, on AWS you could use their load balancer but you could run into various resource limits out of your control there, and pretty much any hardware load balancer would work as well (though this is difficult if not impossible on AWS).
Which load balancer to use is dependent on your (or your personnel's) comfort and knowledge level with each option.
I am using redis version 2.8.3. I want to build a redis cluster. But in this cluster there should be multiple master. This means I need multiple nodes that has write access and applying ability to all other nodes.
I could build a cluster with a master and multiple slaves. I just configured slaves redis.conf files and added that ;
slaveof myMasterIp myMasterPort
Thats all. Than I try to write something into db via master. It is replicated to all slaves and I really like it.
But when I try to write via a slave, it told me that slaves have no right to write. After that I just set read-only status of slave in redis.conf file to false. Hence, I could write something into db.
But I realize that, it is not replicated to my master replication so it is not replicated to all other slave neigther.
This means I could'not build an active-active cluster.
I tried to find something whether redis has active-active cluster capability. But I could not find exact answer about it.
Is it available to build active-active cluster with redis?
If it is, How can I do it ?
Thank you!
Redis v2.8.3 does not support multi-master setups. The real question, however, is why do you want to set one up? Put differently, what challenge/problem are you trying to solve?
It looks like the challenge you're trying to solve is how to reduce the network load (more on that below) by eliminating over-the-net reads. Since Redis isn't multi-master (yet), the only way to do it is by setting up each app server with a master and a slave (to the other master) - i.e. grand total of 4 Redis instances (and twice the RAM).
The simple scenario is when each app updates only a mutually-exclusive subset of the database's keys. In that scenario this kind of setup may actually be beneficial (at least in the short term). If, however, both apps can touch all keys or if even just one key is "shared" for writes between the apps, then you'll need to bake locking/conflict resolution/etc... logic into your apps to consolidate local master and slave differences (and that may be a bit of an overkill). In either case, however, you'll end up with too many (i.e. more than 1) Redises, which means more admin effort at the very least.
Also note that by colocating app and database on the same server you're setting yourself for near-certain scalability failure. What will happen when you need more compute resources for your apps or Redis? How will you add yet another app server to the mix?
Which brings me back to the actual problem you are trying to solve - network load. Why exactly is that an issue? Are your apps so throughput-heavy or is the network so thin that you are willing to go to such lengths? Or maybe latency is the issue that you want to resolve? Be the case as it may be, I recommended that you consider a time-proven design instead, namely separating Redis from the apps and putting it on its own resources. True, network will hit you in the face and you'll have to work around/with it (which is what everybody else does). On the other hand, you'll have more flexibility and control over your much simpler setup and that, in my book, is a huge gain.
Redis Enterprise has had this feature for quite a while, but if you are looking for an open source solution KeyDB is a fork with Active Active support (called Active Replica).
Setting it up is just a little more work than standard replication:
Both servers must have "active-replica yes" in their respective configuration files
On server B execute the command "replicaof [A address] [A port]"
Server B will drop its database and load server A's dataset
On server A execute the command "replicaof [B address] [B port]"
Server A will drop its database and load server B's dataset (including the data it just transferred in the prior step)
Both servers will now propagate writes to each other. You can test this by writing to a key on Server A and ensuring it is visible on B and vice versa.
https://github.com/JohnSully/KeyDB/wiki/KeyDB-(Redis-Fork):-Active-Replica-Support
I have a question that is bugging me quite heavily. What is the Redis pub/sub feature actually used for? I can only think of inter-process communication over TCP (either locally or distributed), however not much else.
Can someone please prove me wrong.
It's an easy way to plug into an event stream, generally between processes or machines. For instance, an user creates a published event. One process handles updating the database from the event, another updates user stats, another global stats, another updates the text search database, etc. They're all loosely coupled by subscribing to the channel. You can add new processes for testing updates and monitoring the system. It's a little different from a message queue in that there's no storing messages until they're processed, but Redis has other structures for those sorts of jobs.
a real use case in my experience.
Lets say you have a web application deployed on 4 different servers(nodes,virtual machines) mostly on your virtual private cloud.
The web application maintains an in memory java map for its static data cache which occasionally changes .
Now every time the data changes in your database you would need all your servers to update there own in memory caches,this is the problem.
one way is to maintain all the static data in redis or any other cache on a separate server and the cache updates based on a scheduler.But here to access the static content which occasionally changes you need a scheduler and a separate cache server like redis or memcached etc. and each server points to this external cache.
Using pubsub of redis here:
all servers subscribe to redis channel and if redis publishes the message when ever there is an update,addition,deletion of the data as a message to all of its subscribers.On receiving the message object and its type of update(ADD,REMOVED,UPDATED) each server updates its in memory static data map.