Assume I have N servers, each operating as a web server and a mongodb member of a replica set.
I'd like the slaveOk reads to be satisfied first by the local mongodb instance, rather than a remote machine across the network.
The documentation says slaveOk reads are satisfied by an arbitrary member. Is it possible to override that?
Mongodb 1.8, C-sharp driver 1.2.
The documentation says slaveOk reads are satisfied by an arbitrary member. Is it possible to override that?
Not without changing the C# driver. You'd probably have to look somewhere in this file to make those changes.
Assume I have N servers, each operating as a web server and a mongodb member of a replica set.
As a note, this is generally not the expected usage for MongoDB. Implemented in this way, your web server will be competing for RAM with MongoDB. If a server gets overloaded the web server will starve the mongod process which will cause connections to back up and exacerbate the issue.
It sounds like you're trying to use MongoDB as a local cache and there are far better tools for this job.
The closest you could come to what you are describing is for each web application to open a separate direct connection (not in replica set mode) to the local mongodb and use that separate connection for reads.
Related
I am using Robomongo tool to connect to various mongo instances. One blocker I faced was that Robomongo do not allow setting readpreference for the particular connection. However, I saw that we can specify Robomongo to load .mongorc.js file at start up. So I added the line rs.slaveOk() to it and now I can connect to slave instances as well.
But this also means that readpreference will be set to secondary even when I am connecting to master instances. I would like to know if there are any disadvantages of keeping this on for all connections. I am sure there is some reason behind Robomongo developers not allowing this as default setting (although setting connection preference for each connection would have been the best solution)
It could be because of stale read issue mentioned here -
[http://docs.mongodb.org/manual/core/read-preference/#edge-cases-2-primaries]
It requires awareness of your replication strategy and 'staleness' of data. The link above is a great read regardless.
edit: answer lacked a brief summary from MongoDB docs linked above ( as mentioned by user 200-ok) "Exercise care when specifying read preferences: Modes other than primary may return stale data because with asynchronous replication, data in the secondary may not reflect the most recent write operations"
I've created a small MongoDB v2.6.4 cluster with one mongos router, three config servers, and handful of mongod data servers. It's working fine, I can do CRUD operations just fine from the mongos client, which directs to the appropriate shards.
I'm now wanting to experiment with the rest service.
When I tried this with a single MongoDB data server using mongod, it worked exactly as prescribed.
Even now, I can connect to each shard which is running with --rest and get a web response. But, the results are a little unexpected. I see the overview, clients, dbtop, write lock, log, etc. And if I click on listDatabase, I actually see my database listed. However, any REST attempts I try return with a JSON object showing offest 0, no rows, total of 0 roes, no query, and 0 milliseconds.
At the moment I'm of the opinion that this doesn't matter, as I've directly connected to a shard and would anticipate only seeing just the contents of that shard -- although further experimentation has been unable to produce even that result.
Again, I'll stress that in the current sharded configuration, mongos, NodeJS, and PHP all can see, query, and manipulate the data within my MongoDB collection just fine.
However, what I'm trying to do (if it's even possible) is connect to the mongos shard service via REST, since it does offer the --httpinterface option, figuring it will expose access to the shards and do the right thing.
It delivers the standard status page, but clicking on any links or hand rolling any REST results in a message about --rest not being enabled and that I should enable it.
REST is not enabled, use --rest to turn on.
check that port 28017 is secured for the network too.
Problem is, mongos doesn't allow that as an option! (mongod, however does)
It is entirely possible this is correct behavior, but I'm not sure how REST services would then work in a sharded configuration.
I'm not using a configuration file, specifying everything by the command line. That said, I tried passing a config file with RESTInterfaceEnabled: true in the yaml, and it didn't help.
mongos is returning a web interface, with the List all commands, db version, and log. It's missing the other diagnostic information (that's usually mongod specific), which I'd perhaps expect. But the links it offers doesn't work -- all requesting me to turn on an option that isn't there.
I've manually checked each shard data server (running mongod with --shardsvr), and directly connecting to them does provide REST access.
Everything else about the sharded cluster is working perfectly. I just can't to REST, like I can in a single-node unsharded solution and access my database or collections.
What might I have missed? Is this even possible? Any ideas?
AFAIK the REST Interface is inly to do monitoring/management not to access the documents. MongoDB does not provide official REST interface to do CRUD operations. For this you need either to create your own or find a library in the ecosystem to do this, look at http://docs.mongodb.org/ecosystem/tools/http-interfaces/
Do you have specific requirements/needs in mind?
PS : I do not use the --rest parameter in any of my deployment/development.
One of our clients have a server running a MongoDB instance and we have to build an analytical application using the data stored in their MongoDB database which changes frequently.
Clients requirements are:
That we do not connect to their MongoDB instance directly or run another instance of MongoDB on their server but just somehow run our own MongoDB instance on our machine in our office using their MongoDB database directory with read only access remotely.
We've suggested deploying a REST application, getting a copy of their database dump but they did not want that. They just want us to run our own MongoDB intance which is hooked up with the MongoDB instance directory. Is this even possible ?
I've been searching for a solution for the past two days and we have to submit a solution by Monday. I really need some help.
I think this is normal request because analytical queries could cause too much load on the production server. It is pretty normal to separate production and analytical databases.
The easiest option is to use MongoDB replication. Set up MongoDB replica set with production database instance as primary and analytical database instance as secondary, also configure the analytical instance to never become primary.
If it is not possible to use replication - for example client doesn't want this, the servers could not connect directly to each other... - there is another option. You can read oplog from remote database and apply operations to your database instance. This is exactly the low level mechanism how replica set works, but you can do it manually too. For example MMS (Mongo Monitoring Sevice) Backup uses reading oplog for online backups of MongoDB.
Update: mongooplog could be the right tool for real-time application of replication oplog pulled from remote server on local server.
I don't think that running two databases that points to the same database files is possible or even recommended.
You could use mongorestore to restore from their data files directly, but this will only work if their mongod instance is not running (because mongorestore will need to lock the directory).
Another solution will be to do file system snapshots and then restore to your local database.
The downside to this backup/restore solutions is that your data will not be synced all the time.
Probably the best solution will be to use replica sets with hidden members.
You can create a replica set with just two members:
Primary - this will be the client server.
Secondary - hidden, with votes and priority set to 0. This will be your local instance.
Their server will always be primary (because hidden members cannot become primaries). Clients cannot see hidden members so for all intents and purposes your server will be read only.
Another upside to this is that the MongoDB replication will do all the "heavy" work of syncing the data between servers and your instance will always have the latest data.
We are looking at CQRS and are evaluating MongoDB as our query database. We have several branch offices with very poor internet connections (256 - 768 mbs DSL) that cannot be upgraded (remote rural locations with no ISP options). The thought was that we could put a box running MongoDB in each branch office and use some sort of clustering so that reads would become very fast since they were accessed over the LAN.
I do not need fail over or high availability.
How do I setup MongoDB so that computers in the branch office talk to the Mongodb instance in the branch office? Note this will be read only.
If MongoDB cannot support this workflow then I am open to suggestions for alternate document based databases that can support this.
You can add a secondary to the MongoDB replica set for each geographical location. Then you can connect to each secondary directly by specifying the hostname of the mongo instance that's local to where you are connecting from, and then issue the slaveOk() command to allow it to be queried. Replication will keep the secondaries in sync with the primary as closely as possible.
Writes/updates will still need to be handled by the primary of the replica set, however.
CouchDB was built for this: occasionally connected document databases.
You can copy a database to your laptop, take it on a plane, modify it there and when you'll have network again, you'll be able to synchronize it with master database. The same works for remote offices.
I strongly suggest that you check it out.
I have used NORM driver in production. New year holidays - it is pretty cool, so my project get high loading and i want to set up a replication set, but have a problem - Norm does not support replication set :( . as far as i understand sharding too?
Help me :) Who did use mongodb csharp or official 10gen driver with replset? Is there any problem on production? If i choose another driver I'll have to rewrite the repository, but I do not want it to be in vain. Is there some issues?
Sharding should not depend on driver-specific support. When you shard, you connect to a router application mongos and this router behaves exactly like mongod.
So you should be able to shard. But you will probably need to change the "connection string". The suggested setup is to have one mongos per application server (instead of your current single mongod).