Log file Viewer
#mongodb logs for Wednesday the 10th of July, 2013
[00:51:26] <charity> hey, i'm trying to create an index on a secondary, and it's not letting me. it says "not master".
[00:51:50] <charity> the command is "c.ensureIndex({"channels": 1, "_created_at": 1}, { background: false})"
[00:52:09] <charity> is there a trick to this? i thought creating indexes on a secondary was the recommended practice.
[01:07:05] <joshua> charity: When we had to do emergency index maint one time 10gen had us change the mongod settings to remove it from the replica set and then add the index
[01:34:53] <joshua> charity: Basically we changed the port so the other servers wouldnt connect, disable the replica set option so it starts as a single mongod and then modify the data
[02:57:32] <EricL> If I create a hashed index (as in {foo: "hashed"}, do I get the same benefit as if I create an index of {foo: 1}?
[04:06:52] <mmlac-bv> Can any one reproduce this? mongo shell does NOT use indexes when used like this: db.foo.find( {"$query":{…..}, "$orderby":{…}) but does use the indexes when used like this db.foo.find({…}).sort{…})?!
[05:54:07] <RoboTeddy> is there a difference between $set: {"key1": {"key2": value}} and $set: {"key1.key2": value}?
[06:40:00] <jkbbwr> My boss asked me to de-normalize a load of sales data and store it in just two tables, for querying later for sales analysis and reporting. Could NoSQL specifically mongodb provide a better solution?
[06:41:40] <jkbbwr> cofeineSunshine: but could it do it well, they want to have an entire KPI platform over the top
[06:44:59] <jkbbwr> I hinted at the idea of it, and he said it might be too complicated, or not the right tool but to research it anyway
[08:11:21] <jonjon> how to change a value in a set? for example, change 'b' to 'c' in this document: {"to" : [ "a", "b" ], "winner" : "0" }
[08:14:29] <jgiorgi> using several collections for queue processing, i'm parallel processing so often find_one will return the same item and two tasks get started on a single queue entry, is it possible for mongo to cycle through and not return the same document on subsequent find_one calls (sharing a cursor is not an option)
[08:17:15] <Nodex> jgiorgi : you will need to implemenent some form of locking of your own, look at findAndModify
[08:17:55] <jgiorgi> Nodex, fair enough, i have a somewhat clumsy method in mind but there's always a way
[08:20:26] <Nodex> when I use Mongo for queue's I lock the document when I query it and unlock it when I'm finished and only select unlocked records
[09:11:17] <schlitzer|work> is here anyone with experience in managing replica sets shards & config servers with puppet?
[09:12:20] <schlitzer|work> doing so in a single host environment(everything on one host for development) is quite easy.
[09:12:52] <schlitzer|work> but i have no clue how one could to this with a real cluster using puppet
[09:44:37] <nyov> I'm looking for an update() query/'spec' that will never match, so an upsert is guaranteed to be an insert. Is there something magic (lightweight) I could use?
[09:59:47] <algernon> Nodex: because upsert allows flat structures, that get turned into structured data
[10:00:22] <algernon> that is, I can get away with {"foo.bar": "baz"} with an upsert, but I need {"foo": {"bar": "baz"}} for insert
[10:01:42] <appi_uppi> I have followed the steps to deploy replica set in our env (Primary, Secondary and an Arbiter) provided in the link http://docs.mongodb.org/manual/tutorial/deploy-replica-set/ ..... but when i run mongodb -f /usr/local/mongodb.conf.. i will get """"forked process: 10914 child process started successfully, parent exiting"
[10:02:28] <nyov> algernon: thanks, but is "_id" not 'more likely' to match some uuid than checking for an 'uuid' key?
[10:08:01] <idank> is it possible to return _all_ subdocuments in an array that match a criteria? $elemMatch only returns the first one
[10:09:37] <Nodex> queries return the parent document by default so everything should be in there anyway
[10:11:04] <idank> huh? $elemMatch can be used to select a specific subdocument from an array, I want all subdocuments
[10:13:32] <nyov> Nodex: sorry, basically I wrap the update query in pymongo, and have already checked the document does not exist (find) and at this stage need to make sure the real upsert doesn't match something that appeared in the meantime, and because of what algernon said I can't run an insert there.
[10:17:41] <Nodex> idank : then perhaps state that in the first place instead of assuming that we know what you mean
[10:20:51] <appi_uppi> Nodex: what could be the reason of getting the message "child process started successfully, parent exiting"?
[12:39:37] <oxsav> I'm trying to do a query to my database, I want to find one specific name in a JsonArray from a collection
[12:42:04] <ron> oxsav: it sounds like you wanna try http://docs.mongodb.org/manual/tutorial/getting-started/ though.
[12:54:47] <Infin1ty> I am using 2.2.5 and i took down one of the replicaset members in order to resync it, i deleted all the data files, started the mongod process again, a few minutes after, the machine is not responsive to anything, i can't ssh, tried to connect using ipmi (it won't pass the login), i do see the replication working using mongostat from other machine, it's resyncing, but it looks like something
[14:09:58] <EricL> If I create a hashed index (as in {foo: "hashed"}, do I get the same benefit as if I create an index of {foo: 1}?
[14:16:04] <Nodex> I would imagine they're used more for hashing an embedded document to save you having to lookup on every field in the embedded doc
[14:19:46] <EricL> Nodex: I meant does it create a hash value and then a range index similar to a :1 or is it a hashed index?
[16:07:18] <Infin1ty> Running sh.stopBalancer(), it somehow waits on a very old mongos process that is not in use since 1 year ago
[16:11:07] <veetow> would i ever use mongos and a config server for a connection to a db which is not sharded?
[16:12:28] <trajik210> I'm just now in production with a mongo cluster, so still a novice. But isn't mongos /only/ for sharded setups? So if you aren't sharding you wouldn't have a mongos daemon.
[16:14:12] <trajik210> I could certainly be wrong, but based on your question I don't think you'd use mongos.
[16:14:13] <veetow> and the reason is that, if say a replica set member needed to be swapped out for a different machine
[16:15:01] <veetow> eg, if i had a mongos running on the local application server pointing at config servers which know about the replica set member changes
[16:16:33] <trajik210> We have a virtualized template for a MongoDB Linux instance. We'd deploy a new machine using the template (no data on the machine).
[16:17:22] <trajik210> We'd give the new instance the same IP as the old server3, add the new machine to the replica set config (telling it to synch from a hidden mongo instance) and then cycle the repl config.
[16:18:05] <trajik210> Once the new instance is done synching it would start to receive queries (depending on the connection string used and the read concern used).
[16:19:05] <veetow> i mean there's got to be a good way to do that without mucking with IPs or connection strings
[16:20:08] <trajik210> If I needed to remove a mongod instance and replace it with a new one, I can do it transparently, on a live production environment, without restarting the app (which is Node.js). But, if I wanted to add a new instance, that would require restarts of the Node.js app in order to add a new IP to the connection string.
[16:21:01] <trajik210> This last example would also be transparent as we run Node.js in a load balanced environ. So we'd do a rolling restart on the Node.js cluster.
[18:33:23] <clarkk> guys, can anyone suggest a library for mongodb that exposes an API to the client to allow mongodb queries to be sent and collections returned, please?
[18:35:07] <dash_> pymongo, what am I doing wrong? http://pastebin.com/GJprZ0BU instead of object with data {u'ok': 1.0, u'value': None} is returned, thanks!!
[18:52:50] <dash_> this is example of syntax I've doing wrong https://github.com/mongodb/mongo-python-driver/blob/master/test/test_collection.py#L1267 now everything works properly, thanks
[19:04:45] <Fallout2man> can someone please recommend me a mongoDB admin tool that works on OSX without me having to hand edit unwieldy config files to point it to my Mongo instance and that doesn't require I install and configure several other pieces of software in a similar way?
[19:05:23] <Fallout2man> Mongo Hub seemed good...but their site is down and I can't get to any binary releases. Just the source repository.
[20:03:34] <nyov> I've been trying to convert a mongodb into a standalone replicaset member, to get the oplog for external fulltext indexing. It seems to work ok, until I restart the mongodb - then it wants to sync with some peers.
[20:04:37] <nyov> None of the sites talking about this setup mention how to disable that, can I force mongo to be primary without some peer?
[20:05:23] <ismell> I want to mapreduce on a ISODate field but I want my granualirty to be a date. How can i strip the time off?
[20:06:25] <trajik210> nyov - are you asking if you can have a single mongo instance be configured in a repl set and be primary (with no other instances in the repl set?)
[20:08:50] <trajik210> Any time I've had a repl set drop to 1 member, it automatically becomes a secondary until at least one other instance comes online. I think you'd have to reconfig the repl set to do what you are saying.
[20:09:15] <nyov> it seemed to work for some dudes here, maybe in an earlier mongodb version? https://loosexaml.wordpress.com/2012/09/03/how-to-get-a-mongodb-oplog-without-a-full-replica-set/
[20:09:47] <trajik210> Could be. I'm using 2.4.3 and 2.4.4. You can't have a primary instance in a repl set with one member.
[20:10:09] <trajik210> You'd likely have to drop the instance you are interested in from the repl set using mongo shell.
[20:11:12] <trajik210> If you only need the oplog for synching purposes could you not just ensure your instance is caught up then drop it from the repl set? It'd effectively become its own instance at that point.
[20:12:28] <nyov> trajik210: nah, I don't need it for syncing. I need the tailable oplog to get updates/inserts/deletes into elasticsearch
[20:21:29] <ismell> I'm doing a mapreduce like so emit(Date.UTC(this.created_at.getUTCFullYear(), this.created_at.getUTCMonth(), this.created_at.getUTCDay()), 1);
[20:22:50] <nyov> http://stackoverflow.com/questions/10413349/get-component-of-date-isodate-in-mongo
[20:23:33] <lucasliendo> Hi everyone ! Have someone built a mongo cluster ? I have successfully built a mongo cluster, and now have some questions that I don't seem to find answers from the docs
[20:25:53] <lucasliendo> my cluster is completly empty, so now I would like to put in there a huge collection (this huge collection is a dump from another mongo, which is running stand-alone)
[20:26:36] <nyov> ismell: isodate is just a helper function. you can use the same methods from a normal JS Date() object
[20:27:28] <lucasliendo> do i have to worry at all for anything but the shard key ? I mean, can I just fill my cluster with mongoimport ? What about the indexes from the original collection (are they lost) ?
[20:29:43] <lucasliendo> sure, I have 6 machines, the cluster is built with 3 replica sets (rs1, rs2 & rs3), and each replica set is using 2 machines
[20:30:31] <lucasliendo> also have 3 mongo config servers among this machines, and only one of them is running mongos
[20:34:21] <trajik210> I don't have a similar setup as you as I'm just repl sets w/o sharding. We use mongoimport to bulk import a lot of documents regularly.
[20:34:31] <lucasliendo> so, basically I would like to know the best way to do this...that's why I asked in first place about the shard key...from mongo docs it seems the only relevant thing...am I missing sometring else ?
[20:35:22] <trajik210> lucasliendo: You should be okay if all you are doing is importing an existing data set into an empty system.
[20:36:11] <trajik210> lucasliendo: You mean if you run multiple mongoimport processes and some of the docs are in both import runs?
[20:37:37] <lucasliendo> no no, just one mongoimport. I mean is this enough to import the whole collection : 1 -> Create an empty db. 2 -> Select the shard key. 3 -> run mongoimport
[20:41:55] <lucasliendo> ok, well...that was my doubt after all... :D. Just one more thing, while you were importing, how did you check that everything was ok ? I mean the import was running ok against the whole cluster...
[20:42:06] <trajik210> lucasliendo: You can use mongotop or mongostat to monitor the instance (from another Terminal) as you are importing.
[21:07:53] <BlakeRG> Hello, i've got a collection that i want to create a unique index on, my document looks like this: { "_id" : ObjectId("51ddc984375e7d25b6000002"), "id" : "98374374948324", "first_name" : "Joe", "last_name" : "Blow" }
[21:07:59] <BlakeRG> db.pooballs.ensureIndex( { id: 1 }, { dropDups: true } ); does not seem to do the trick
[21:17:44] <BlakeRG> crudson1: just figured it out…. db.pooballs.ensureIndex( { id: 1 }, { unique: true, dropDups: true } )
[21:22:54] <crudson1> BlakeRG: The docs there imply that dropdups implies unique. Looks like it needs updating. Seems to be right here: http://docs.mongodb.org/manual/core/indexes/#drop-duplicates
[21:23:34] <BlakeRG> crudson1: yes i just found that a second ago, the tutorial makes you think it's implied