So a large part of my work involves a postgres system spread between many machines. The way we've been handling sharding is to create a separate postgres instance for each database shard. There are about ~10 shards per machine, each is bound to a distinct port, systemctl service, data root, config, etc.
These are large databases which have the potential to consume a lot of memory.
I inherited the sharding system from a previous developer, but I haven't thought to change it because it has been functioning so far. Is this a bad practice? Are there any potential side effects?
It is fine to have multiple PostgreSQL instances on a single machine to be able to start, stop and upgrade databases independently.
However, it is strange to see a sharding solution where multiple shards are on a single machine. That seems like a waste of resources. But perhaps the idea is to move shards to other machines later, when load increases.
Related
We're using amazon web service for a business application which is using node.js server and mongodb as database. Currently the node.js server is runing on a EC2 medium instance. And we're keeping our mongodb database in a separate micro instance. Now we want to deploy replica set in our mongodb database, so that if the mongodb gets locked or unavailble, we still can run our database and get data from it.
So we're trying to keep each member of the replica set in separate instances, so that we can get data from the database even if the instance of the primary memeber shuts down.
Now, I want to add load balancer in the database, so that the database works fine even in huge traffic load at a time. In that case I can read balance the database by adding slaveOK config in the replicaSet. But it'll not load balance the database if there is huge traffic load for write operation in the database.
To solve this problem I got two options till now.
Option 1: I've to shard the database and keep each shard in separate instance. And under each shard there will be a reaplica set in the same instance. But there is a problem, as the shard divides the database in multiple parts, so each shard will not keep same data within it. So if one instance shuts down, we'll not be able to access the data from the shard within that instance.
To solve this problem I'm trying to divide the database in shards and each shard will have a replicaSet in separate instances. So even if one instance shuts down, we'll not face any problem. But if we've 2 shards and each shard has 3 members in the replicaSet then I need 6 aws instances. So I think it's not the optimal solution.
Option 2: We can create a master-master configuration in the mongodb, that means all the database will be primary and all will have read/write access, but I would also like them to auto-sync with each other every so often, so they all end up being clones of each other. And all these primary databases will be in separate instance. But I don't know whether mongodb supports this structure or not.
I've not got any mongodb doc/ blog for this situation. So, please suggest me what should be the best solution for this problem.
This won't be a complete answer by far, there is too many details and I could write an entire essay about this question as could many others however, since I don't have that kind of time to spare, I will add some commentary about what I see.
Now, I want to add load balancer in the database, so that the database works fine even in huge traffic load at a time.
Replica sets are not designed to work like that. If you wish to load balance you might in fact be looking for sharding which will allow you to do this.
Replication is for automatic failover.
In that case I can read balance the database by adding slaveOK config in the replicaSet.
Since, to stay up to date, your members will be getting just as many ops as the primary it seems like this might not help too much.
In reality instead of having one server with many connections queued you have many connections on many servers queueing for stale data since member consistency is eventual, not immediate unlike ACID technologies, however, that being said they are only eventually consistent by 32-odd ms which means they are not lagging enough to give decent throughput if the primary is loaded.
Since reads ARE concurrent you will get the same speed whether you are reading from the primary or secondary. I suppose you could delay a slave to create a pause of OPs but that would bring back massively stale data in return.
Not to mention that MongoDB is not multi-master as such you can only write to one node a time makes slaveOK not the most useful setting in the world any more and I have seen numerous times where 10gen themselves recommend you use sharding over this setting.
Option 2: We can create a master-master configuration in the mongodb,
This would require you own coding. At which point you may want to consider actually using a database that supports http://en.wikipedia.org/wiki/Multi-master_replication
This is since the speed you are looking for is most likely in fact in writes not reads as I discussed above.
Option 1: I've to shard the database and keep each shard in separate instance.
This is the recommended way but you have found the caveat with it. This is unfortunately something that remains unsolved that multi-master replication is supposed to solve, however, multi-master replication does add its own ship of plague rats to Europe itself and I would strongly recommend you do some serious research before you think as to whether MongoDB cannot currently service your needs.
You might be worrying about nothing really since the fsync queue is designed to deal with the IO bottleneck slowing down your writes as it would in SQL and reads are concurrent so if you plan your schema and working set right you should be able to get a massive amount of OPs.
There is in fact a linked question around here from a 10gen employee that is very good to read: https://stackoverflow.com/a/17459488/383478 and it shows just how much throughput MongoDB can achieve under load.
It will grow soon with the new document level locking that is already in dev branch.
Option 1 is the recommended way as pointed out by #Sammaye but you would not need 6 instances and can manage it with 4 instances.
Assuming you need below configuration.
2 shards (S1, S2)
1 copy for each shard (Replica set secondary) (RS1, RS2)
1 Arbiter for each shard (RA1, RA2)
You could then divide your server configuration like below.
Instance 1 : Runs : S1 (Primary Node)
Instance 2 : Runs : S2 (Primary Node)
Instance 3 : Runs : RS1 (Secondary Node S1) and RA2 (Arbiter Node S2)
Instance 4 : Runs : RS2 (Secondary Node S2) and RA1 (Arbiter Node S1)
You could run arbiter nodes along with your secondary nodes which would help you in election during fail-overs.
Replication seems to be a lot simpler than sharding, unless I am missing the benefits of what sharding is actually trying to achieve. Don't they both provide horizontal scaling?
In the context of scaling MongoDB:
replication creates additional copies of the data and allows for automatic failover to another node. Replication may help with horizontal scaling of reads if you are OK to read data that potentially isn't the latest.
sharding allows for horizontal scaling of data writes by partitioning data across multiple servers using a shard key. It's important to choose a good shard key. For example, a poor choice of shard key could lead to "hot spots" of data only being written on a single shard.
A sharded environment does add more complexity because MongoDB now has to manage distributing data and requests between shards -- additional configuration and routing processes are added to manage those aspects.
Replication and sharding are typically combined to created a sharded cluster where each shard is supported by a replica set.
From a client application point of view you also have some control in relation to the replication/sharding interaction, in particular:
Read preferences
Write concerns
Consider you have a great music collection on your hard disk, you store the music in logical order based on year of release in different folders.
You are concerned that your collection will be lost if drive fails.
So you get a new disk and occasionally copy the entire collection keeping the same folder structure.
Sharding >> Keeping your music files in different folders
Replication >> Syncing your collection to other drives
Replication is a mostly traditional master/slave setup, data is synced to backup members and if the primary fails one of them can take its place. It is a reasonably simple tool. It's primarily meant for redundancy, although you can scale reads by adding replica set members. That's a little complicated, but works very well for some apps.
Sharding sits on top of replication, usually. "Shards" in MongoDB are just replica sets with something called a "router" in front of them. Your application will connect to the router, issue queries, and it will decide which replica set (shard) to forward things on to. It's significantly more complex than a single replica set because you have the router and config servers to deal with (these keep track of what data is stored where).
If you want to scale Mongo horizontally, you'd shard. 10gen likes to call the router/config server setup auto-sharding. It's possible to do a more ghetto form of sharding where you have the app decide which DB to write to as well.
Sharding
Sharding is a technique of splitting up a large collection amongst multiple servers. When we shard, we deploy multiple mongod servers. And in the front, mongos which is a router. The application talks to this router. This router then talks to various servers, the mongods. The application and the mongos are usually co-located on the same server. We can have multiple mongos services running on the same machine. It's also recommended to keep set of multiple mongods (together called replica set), instead of one single mongod on each server. A replica set keeps the data in sync across several different instances so that if one of them goes down, we won't lose any data. Logically, each replica set can be seen as a shard. It's transparent to the application, the way MongoDB chooses to shard is we choose a shard key.
Assume, for student collection we have stdt_id as the shard key or it could be a compound key. And the mongos server, it's a range based system. So based on the stdt_id that we send as the shard key, it'll send the request to the right mongod instance.
So, what do we need to really know as a developer?
insert must include a shard key, so if it's a multi-parted shard key, we must include the entire shard key
we've to understand what the shard key is on collection itself
for an update, remove, find - if mongos is not given a shard key - then it's going to have to broadcast the request to all the different shards that cover the collection.
for an update - if we don't specify the entire shard key, we have to make it a multi update so that it knows that it needs to broadcast it
Whenever you're thinking about sharding or replication, you need to think in the context of writers/update operations. If you don't need to scale writes then replications, as it fairly simpler, is a good choice for you.
On the other hand, if you workload mostly updates/writes then at some point you'll hit a write bottleneck. If write request comes Mongo blocks other writes request. Those write request blocks until the first request will be done. If you want to scale this writes and want parallelize it then you need to implement sharding.
Just to put this somewhere...
The most basic way to run mongo is as standalone server.
You write a config (file or cli options)
initiate the server using mongod
For this picture, I didn't include the "client". Check the next one.
A replica set is a set of servers initialized exactly as above with a different config file.
To link them, we connect to one of them, and initialize the replica set mode.
They will mirror each other (in the most common configuration). This system guarantees high availability of data.
The initialization of the replica set is represented in the red border box.
Sharding is not about replicating data, but about fragmenting data.
Each fragment of data is called chunk and goes to a different shard. shard = each replica set.
"main" server, running mongos instead of mongod. This is a router for queries from the client.
Obvious: The trade-off is a more complex architecture.
Novelty: configuration server (again, a different config file).
There is much more to add, but apart from the words the pictures hold much the same.
Even mongoDB recommends to study your case carefully before going sharding. Vertical scaling (vs) is probably a good idea at least once before horizontal scaling (hs).
vs is done upgrading hardware (cpu, ram, etc). hs is needs more computers (but could be cheap computers).
Both replication and sharding can be used (individually or together) for horizontal scaling of a MongoDB installation.
Sharding is MongoDB's solution for meeting the demands of data growth. Sharding stores data records across multiple servers to provide faster throughput on read and write queries, particularly for very large data sets.
Any of the servers in the sharded cluster can respond to a read or write operation, which greatly speeds up query responses.
Replication is MongoDB's solution for providing stability, backup, and disaster recovery to a MongoDB installation. This process copies and synchronizes the replica data set across multiple servers. This prevents downtime if one server goes offline.
Any of the secondary servers can respond to read queries, but only the primary server will perform write operations. The results of the write operation will then be propagated out to the secondary servers.
Scenario 1: Fault-Tolerance
In this scenario, the user is storing billing data in a MongoDB installation. This data is mission-critical to the user's business, and needs to be available 24/7, even if a server crashes or is taken offline.
MongoDB replication is the best solution for this user. With replication, the entire data set is mirrored on multiple servers. If a server fails or is taken offline, the other servers in the cluster take over.
Scenario 2: High Performance
In this scenario, the user is running a social networking site which is run from a MongoDB database. As the social network grows, the MongoDB data set has grown along with it. The user is seeing query times and page loads increase beyond an acceptable point. It is critical that the user's MongoDB installation receives a major performance boost.
Setting up a sharded MongoDB cluster is the best solution for this user. The sharded cluster will break up the user's data set and store parts of it on separate secondary servers. Each secondary server can respond to read or write queries on its portion of the data, which greatly increases the installation's response time
MongoDB Atlas is a Database as a service in could. It support three major cloud providers such as Azure , AWS and GCP. In cloud environment , we usually talk about high availability and scalability. In Atlas “clusters”, can be either a replica set or a sharded cluster.
These two address high availability and scalability features of our cloud environment.
In general Cluster is a group of servers used to achieve a specific task. So sharded clusters are used to store data in across multiple machines to meet the demand of data growth. As the size of the data increases, a single machine may not be sufficient to store the data nor provide an acceptable read and write throughput. Sharded clusters supports the horizontal scalability of the underling cloud environment.
A replica set in MongoDB is a group of mongod processes that maintain the same data set. Replica sets provide redundancy and high availability, and are the basis for all production deployments.In a replica, one node is a primary node that receives all write operations. All other instances, such as secondaries, apply operations from the primary so that they have the same data set. Replica set mainly focus on the availability of data.
Please check the documentation
Thank You.
As far as I understand sharding (e.g in MongoDB) and distributed file systems (e.g. HDFS in HBase or HyperTable) are different mechanisms that databases use to scale-out, however I wonder how do they compare?
Traditional sharding involves breaking tables into a small number of pieces and running each piece (or "shard") in a separate database on a separate machine. Because of the large shard size, this mechanism can be prone to imbalances due to hot spots and unequal growth as was evidenced by the Foursquare incident. Also, because each shard is run on a separate machine, these systems can experience availability problems if one of the machines goes down. To mitigate this problem, most sharding systems, including MongoDB, implement replica groups. Each machine is replaced by a set of three machines in a master plus two slaves configuration. This way if a machine goes down, there are two remaining replicas to serve the data. There are a couple of problems with this design: First, if a replica fails in a replica group, and the group is only left with two members, to bring the replication count back to three, the data on one of these two machines needs to be cloned. Since there are only two machines in the entire cluster that can be used to re-create the replica, there will be enormous drag on one of these two machines while re-replication is taking place, causing serious performance problems on the shard in question (it takes over two hours to copy 1TB over a gigabit link). The second problem is that when one of the replicas goes down, it needs to be replaced with a new machine. Even if there is plenty of spare capacity across the cluster to resolve the replication problem, that spare capacity cannot be used to rectify the situation. The only way to solve it is to replace the machine. This becomes very challenging from an operational standpoint as cluster sizes grow up into the hundreds or thousands of machines.
The Bigtable+GFS design solves these problems. First, the table data is broken down into much finer grained "tablets". A typical machine in a Bigtable cluster will often have 500+ tablets. If an imbalance occurs, resolving it is just a simple matter of migrating a small number of tablets from one machine to another. If a TabletServer goes down, because the data set is broken down and replicated with such fine granularity, there can be hundreds of machines that participate in the recovery process, which distributes the recovery burden and speeds recovery time. Also, because the data is not tied to a specific machine or machines, the spare capacity on all machines in the cluster can be applied to the failure. There is no operational requirement to replace the machine since any of the spare capacity throughout the cluster can be used to rectify replication imbalance.
Doug Judd
CEO, Hypertable Inc.
I am currently evaluating MongoDb for our write heavy application...
Currently MongoDb uses single thread for write operation and also uses global lock whenever it is doing the write... Is it possible to exploit multiple CPU on a multi-CPU server to get better write performance? What are your workaround for global write lock?
No, it is still recommended to use sharding to utilize multiple CPU cores.
As stated in the FAQ
Sharding improves concurrency by distributing collections over multiple mongod instances, allowing shard servers (i.e. mongos processes) to perform any number of operations concurrently to the various downstream mongod instances.
Each mongod instance is independent of the others in the shard cluster and uses the MongoDB readers-writer lock). The operations on one mongod instance do not block the operations on any others.
Sharding on a single box has its issues, as one user stated in the mongodb-user mailing list
After some significant experimentation, I've found a single MongoDB shard daemon CANNOT use more than one CPU. On a 24 CPU box, performance scales up until we hit about 8 shards, then another limit kicks in.
So right now, the easy solution is to shard.
Yes, normally sharding is done across servers. However, it is completely possible to shard on a single box. You simply fire up the shards on different ports and provide them with different folders. Here's a sample configuration of 2 shards on one box.
The MongoDB team recognizes that this is kind of sub-par, and I know from talking to them that they're looking at better ways to do this.
Obviously once you get multiple shards on one box and increase your write threads, you will have to be wary of disk IO. In my experience, I've been able to saturate disks with a single write thread. If your inserts/updates are relatively simple, you may find that extra write threads don't do anything. (Map-Reduces are the exception here, sharding definitely helps there)
Consider the following setup:
There a 2 physical servers which are set up as a regular mongodb replication set (including an arbiter process, so automatic failover will work correctly).
now, as far as i understand, most actual work will be done on the primary server, while the slave will mostly just do work to keep its dataset in sync.
Would it be reasonable, to introduce sharding into this setup in a way that one would set up another replication set on the same 2 servers, so that each of them has one mongod process running as primary and one process running as secondary.
The expected result would be that both servers will share the workload of actual querys/inserts while both are up. In the case of one server failing the whole setup should elegantly fail over to continue running, until the other server is restored.
Are there any downsides to this setup, except the overall overhead in setup and number of processes (mongos/configservers/arbiters)?
That would definitely work. I'd asked a question in the #mongodb IRC channel a bit ago as to whether or not it was a bad idea to run multiple mongod processes on a single machine. The answer was "as long as you have the RAM/CPU/bandwidth, go nuts".
It's worth noting that if you're looking for high-performance reads, and don't mind writes being a bit slower, you could:
Do your writes in "safe mode", where the write doesn't return until it's been propagated to N servers (in this case, where N is the number of servers in the replica set, so all of them)
Set the driver-appropriate flag in your connection code to allow reading from slaves.
This would get you a clustered setup similar to MySQL - write once on the master, but any of the slaves is eligible for a read. In a circumstance where you have many more reads than writes (say, an order of magnitude), this may be higher performance, but I don't know how it'd behave when a node goes down (since writes may stall trying to write to 3 nodes, but only 2 are up, etc - that would need testing).
One thing to note is that while both machines are up, your queries are being split between them. When one goes down, all queries will go to the remaining machine thus doubling the demands placed on it. You'd have to make sure your machines could withstand a sudden doubling of queries.
In that situation, I'd reconsider sharding in the first place, and just make it an un-sharded replica set of 2 machines (+1 arbiter).
You are missing one crucial detail: if you have a sharded setup with two physical nodes only, if one dies, all your data is gone. This is because you don't have any redundancy below the sharding layer (the recommended way is that each shard is composed of a replica set).
What you said about the replica set however is true: you can run it on two shared-nothing nodes and have an additional arbiter. However, the recommended setup would be 3 nodes: one primary and two secondaries.
http://www.markus-gattol.name/ws/mongodb.html#do_i_need_an_arbiter