Log file Viewer
#mongodb logs for Thursday the 13th of September, 2012
[00:13:02] <gustonegro> _m: yeah, it only creates objects. I do: db.things.update({_id: ObjectId("5048ccc355d87a0178000003") }, { "$set" : { "content.foo.3.bar" : "5" } } )
[00:13:12] <gustonegro> and I get: { "__v" : 0, "_id" : ObjectId("5048ccc355d87a0178000003"), "content" : { "foo" : { "3" : { "bar" : "5" } } } }
[00:13:51] <gustonegro> is there no way in dot notation to specify an array? I tried "foo[4].bar", but it just creates a field called "foo[4]"
[00:20:11] <_m> gustonegro: Have you tried something like db.things.update({_id: ObjectId('123asd'), 'content' : { 'foo' : { '$' : {'bar' : 5 } } } })
[00:25:51] <gustonegro> _m: when I fix up the objects with: db.things.update({_id: ObjectId("5048ccc355d87a0178000003")}, {'content' : { 'foo' : { '$' : {'bar' : 5 } } } })
[00:27:03] <gustonegro> if I do: db.things.update({_id: ObjectId("5048ccc355d87a0178000004")}, {'content' : { 'foo' : { 4 : {'bar' : 5 } } } })
[01:09:59] <gustonegro> _m: yeah, that works, but is like you say verbose....in which case I wouldn't need dot notation
[01:10:59] <gustonegro> I would _assume_ that dotnotation using an update with { $set : { 'content.foo.0' : {bar: 8} } } would create and array at "foo" and not an object (if the object or array doesn't already exist)
[01:12:02] <gustonegro> even better, I think would be if dot notation had a concept of array ...like: content.foo[0]
[01:25:00] <gustonegro> _m: no worries...thanks for the help! I'm guessing it's just not possible what I want
[01:31:02] <tomlikestorock> I'm trying to add a new secondary to my mongo replica set, and I ran rs.initiate and now I have two primaries. How do set the new node as a secondary?
[02:09:52] <ejcweb> Is it possible to search for documents where one field is greater than another, for example?
[02:14:15] <crudson> ejcweb: you can do this with http://docs.mongodb.org/manual/reference/aggregation/#aggregation-pipeline-operator-reference
[02:23:12] <ejcweb> Thanks crudson. Are you able to point me towards a specific example for comparing fields of the same document?
[02:29:08] <tomlikestorock> what php driver version do I need to connect to mongo2.2? I keep getting "could not determine master" even though all the mongos say there's a master (and it's reported as the same on all of them)
[02:45:56] <crudson> ejcweb: oh I don't know anything about it. A quick search showed a mongo hosting company. They don't give you a shell to use?
[08:15:33] <Gargoyle> could I then query and fetch docs using find({newsletters: {type: B2B, enabled: true}})
[08:16:41] <crudson> Gargoyle: yes, but ensure you use $elemMatch and dot notation for the attributes rather than a subdocument as in your example
[08:24:00] <Gargoyle> NodeX: Not had chance to look at it. But will be interested in your dataset though.
[08:26:04] <NodeX> it has 41k placenames, 1.8m full postcodes, all of the part (first part) of the postcodes, I can probably add counties in there for you too
[08:31:46] <crudson> Gargoyle: my mistake, you don't need dot notation. this will work for you: db.garg.find({newsletters:{$elemMatch:{type:'B2B',enabled:true}}})
[08:58:38] <arussel> I'm trying to find how to do "select count(*) from foo", I 've tried db.foo.size(), db.foo.length() and db.foo.count() without success. Is there a way ?
[10:18:35] <driek> anyone using mongodb from aptitude did an upgrade to 2.2 already? does the recommended way of upgrading work well in an aptitude scenario?
[10:27:01] <driek> in the setup I have to upgrade, daemontools is used instead of sysinit. But I think the mongod's will keep running (the old version) if I only shut down mongos, do the upgrade, and restart mongos
[10:29:15] <marcqualie> mine worked fine, but I'm not using shards so mongos didn't play any part in the upgrade
[11:22:35] <skieter> Following EC2 Quickstart on mongodb.org: Can anyone tell me why you'd set up RAID10 rather than just RAID0 - EBS is supposed to already have redundancy?
[11:37:31] <ron> skieter: http://serverfault.com/questions/295324/why-mongodb-docs-recommends-to-run-raid-10-on-ebs
[11:47:45] <falu_> i've got one update() call with pymongo that works on my local machine but not on my server. i can insert() a copy of a document (that isn't being updated on the server) on my local machine and there the same update() call will work. both run the same mongodb version.i have no idea how to find the cause.
[11:56:00] <skieter> balboah: yes, this article seems to refer to some benchmarks that actually show RAID10 to be much slower than RAID0, even thought it does not make sense http://www.nevdull.com/2008/08/24/why-raid-10-doesnt-help-on-ebs/
[12:15:37] <balboah> skieter: also perhaps the "ebs optimized" option on amazon came after those articles
[12:18:22] <falu_> found the cause for my problem above. i had changed indexes in my pymongo code and the server database didn't care and is still using the old index settings.
[12:29:28] <algernon> moses: http://www.mongodb.org/display/DOCS/Advanced+Queries#AdvancedQueries-%24or
[12:57:30] <jmar777> matubaum: you shouldn't need to use the ordinal value - that's mostly used internally to insure that all generated Timestamps are unique (within the same mongod process, anyway)
[13:03:13] <matubaum> jmar777: thanks , but how can I create a timestamp to include it inside a query to ask for documents with "created" less than this timestamp?
[13:03:14] <matubaum> I was doing it like this, but it doesn't work: db.cacheproduct.find( {"created": { "$lt": new Timestamp("1347538548001", 1) }} )
[13:05:24] <jmar777> matubaum: hm... i'm honestly not that familiar with it. the docs specify a little behavior around them, but it's considered an internal data type: http://www.mongodb.org/display/DOCS/Timestamp+data+type
[13:07:27] <matubaum> jmar777: yeap, but that doc doesn't say to much, it doesn't say anything about Timestamp Api
[13:08:52] <jmar777> matubaum: seems like an issue with the ODM then - maybe you should define your type as a number rather than a Timestamp
[13:09:25] <jmar777> matubaum: i'm surprised it wouldn't use a Date object by default - that's the supported date/time API in mongo
[13:10:00] <jmar777> matubaum: i think it's risky for you to allow it to use a Timestamp, just because you'll be dealing with a lot of unspecified behavior
[13:32:53] <jcromartie> like I said: it doesn't make any sense for virtual machines, SSDs, and possibly many other cases
[13:33:10] <Gargoyle> I thought it was because it uses memory mapped files, and a certain amount of "working space" is required
[13:36:43] <jcromartie> I know preallocation is valuable for big write-heavy instances on magnetic disks
[13:38:23] <Gargoyle> And unless you are running some custom hardware, then your VM storage is probably on a SAN/iSCSI somewhere ?
[13:44:08] <statusfailed> I started a MapReduce operation, and stupidly tried to kill it with ctrl+c
[14:53:07] <IAD1> jiffe98: mongo use mmap to map files in RAM. it you have much RAM - all data will be in memory. http://www.mongodb.org/display/DOCS/Checking+Server+Memory+Usage#CheckingServerMemoryUsage-MemoryMappedFiles
[14:57:52] <falu_> what happens when my pymongo app which creates/ensures indexes in a collection is running and i drop the index of a collection from the shell and re-create it? will requests while the index is gone still be served?
[15:01:27] <falu_> balboah: hm, slowness would not be a problem. blocking would. i tried it on my dev machine and didn't notice any differences while the index was gone, but i'm not sure if i am overseeing the problem here.
[15:03:58] <falu_> kali: okay, my question was rather what happens while the index is not available? will my app stumble because it thinks there is an index?
[15:04:48] <kali> there is an option you can start mongo with, that will stricly forbid a query to perform a table scan
[15:05:59] <falu_> well, actually i have another question... when i change an option for an index (like enabling sparse) with pymongo, the index won't be changed accordingly without manual intervention. is that correct?
[15:06:43] <kali> last time i checked, with the ruby driver and the javascript shell, you had to drop the index and create it again
[15:07:32] <kali> it makes sense, actualy. what you are trying to do will alter deeply the nature of the index, so...
[15:10:18] <falu_> kali: exactly. that's why i think it could be well reasoned that it must be done manually.
[15:55:40] <gustoneg1o> I've been having serious difficulty trying to get mongodb to update values in a document. I want it to create the value if it doesn't exist, including the entire path up to it.
[17:04:03] <jY> http://www.mongodb.org/display/DOCS/Reconfiguring+a+replica+set+when+members+are+down
[17:43:52] <mandark> Hello world, one question : Is it possible in Python to insert a (key, value) in a SON at a specified position ?
[17:45:44] <Almindor> how would you go about finding all documents which have 2 array fields of which length differs
[17:47:38] <kali> Almindor: or yo ucan using $where but it's using the javascript interpreter, so not for a production request
[17:48:45] <kali> Almindor: and if other bit of your app are using the javascript engine, they wil get stuck
[17:55:15] <mandark> So it's not possible, in Python, to insert a (key, value) pair in a SON object at a specified position ?
[17:57:55] <_m> mandark: Not sure how it would work in Python, but here are some examples in the console
[17:59:55] <mandark> _m: I'm ok with inserting, but as MongoDB's bson is order sensitive, I need a way to create "key ordered documents", it's provided by bon.son.SON in pymongo.
[18:00:13] <mandark> _m: constructing a SON is easy with SON([(key1, value1), (key2, value2), and so on])
[18:00:41] <mandark> _m: but what about editing (adding a key value pair) in an existing SON object at a "not random" position ?
[18:07:58] <mandark> _m: I got an idea, we should build a son_manipulator that sort document keys :-X no more problem about field ordering ...
[18:15:13] <mandark> _m: Yes but as the SON object is not documented I don't know any way to implement this kind of son_manipulator ... :X
[19:41:26] <fitzagard> Any specific reason why the mongo-php-driver on 1.2.12 would throw a depreciated notice for MongoCursor::group() - Implicitly passing condition as $options will be removed in the future?
[19:43:31] <rnickb> how would you write this query using the c++ api? db.records.find({_id : { $gt : {userid: 1, timestamp : 0}, $lte : {userid: 1, timestamp: 100}}})
[20:15:18] <gustonegro> Hi, is there a reason why dot notation doesn't consider numbers to be indexs in arrays? (when upserting or inserting)
[20:45:37] <aboudreault> When we have many documents to get (say 100-1000), and have to do complex operation. (merging, stat). is it better to create *relatively big* mapredude fonctions, or would be better to fetch my document using a script on the server, then process things ?
[21:08:53] <_m> gustonegro: db.things.update({ _id : ObjectId("50512dd15e5110032d1fa00d") }, { $set : { 'content.foo' : [{}, {bar: 10}] } }, true)
[21:25:17] <gustonegro> _m: I tried that too, but then it overwrites everything at "content.foo" ...which is not what I want if I have values in content.foo.0 and only want to update content.foo.1.bar
[21:43:32] <gustonegro> I'm guessing there was some logic as to why upserting/inserting with a dot notation content.foo.1.bar creates an Object at foo instead of an Array, no?
[21:44:22] <gustonegro> even better, I think, would be a more explicit dot notation with content.foo[1].bar
[22:48:56] <aboudreault> When we have many documents to get (say 100-1000), and have to do complex operation. (merging, stat). is it better to create *relatively big* mapredude fonctions, or would be better to fetch my document using a script on the server, then process things ?
[22:58:42] <jiffe98> will it work to have 4 sets of machines, 2 of which have more disk space than the other two and also take advantage of that extra disk space?
[23:07:09] <_m> jiffe98: Mongo itself shouldn't care. Really depends on your architecture. In practice, I've found it easier to have a consistent master/slave machine sizes
[23:08:21] <jiffe98> well I have 2 machines in a mirror right now, and I'm looking to add 2 more machines as a second shard
[23:08:23] <_m> aboudreault: Depends on your resources and usage. I prefer to keep stats processing outside of Mongo in preference to StatsD, Hadoop, etc. as to limit performance implications.
[23:09:02] <_m> That said, it shouldn't matter *too much* if the machine is only there to process stats.
[23:09:41] <aboudreault> _m, in fact, you might be able to recommend me some way to work. If you have 2-5 minutes?
[23:09:58] <jiffe98> so as far as mongodb is concerned both machines are equal so usage-wise it will max out when both sides reach the disk space on the smaller set
[23:11:43] <aboudreault> I want to build something that each user has a kind of bucket and they can upload a document, which has no relation with any other. Each document represent something with date.That's why I'm interested in mongodb. The problem, and the why I'm hesitating to use postgresql, is because I would like to build statistic with these data. So postgreSQL queryability is awesome for that.
[23:12:30] <aboudreault> _m, but I thought that since there is no *performance guarantee* at all and that it is not a high critiria .... I could only use complex mapreduce fonctions
[23:13:08] <aboudreault> most of time, it should be easy and simple. but something (ie, with multiples documents... or even some day with multiples documents of multiple users) it might become harder.
[23:13:42] <aboudreault> you are talking about statD and hadoop. Would those be something I should take a look? Are they a replacement of what I want to push in mongodb, or a complement?
[23:14:50] <wzlwzl> i have 2 mongod +1 arb in a replica set.. i had data corruption… now both mongod are in STARTUP2 and I need to get out of this state to restore… how can I make one of them primary?
[23:29:59] <_m> aboudreault: Depending on your language, an ORM would give you that functionality (joins, essentially)
[23:31:00] <_m> aboudreault: Hadoop is a great tool for deterministic stat processing. It's a map-reduce platform, stated basically.
[23:33:54] <_m> There's a discussion of some Python ORMs. I'm not familiar with python and mongo, so I haven't a recommendation
[23:34:29] <_m> aboudreault: http://hadoop.apache.org/#What+Is+Apache+Hadoop%3F for more information
[23:35:41] <_m> A "users" collection with array field named "docs" with each document's information should work.
[23:36:11] <_m> We do something similar, with the actual files stored in S3 and references to those within an array in our user documents.
[23:41:07] <wzlwzl> "errmsg" : "replSetReconfig command must be sent to the current replica set primary.",
[23:43:53] <wzlwzl> "assertion" : "initiation and reconfiguration of a replica set must be sent to a node that can become primary"
[23:47:51] <_m> wzlwzl: I've never run into this before. Will send our sysops guy a message and see if he can shed some light.
[23:48:25] <_m> Also, you could attempt to shut down the "slave" and see if election happens, then bring the slave back up
[23:49:38] <R-66Y> is there a native mongoDB shell way to figure out how "long" an associative array is?
[23:50:31] <tomlikestorock> anyone know what this error means? [TTLMonitor] problem detected during query over admin.system.indexes : { $err: "not master or secondary; cannot currently read from this replSet member", code: 13436 }
[23:50:45] <tomlikestorock> or this? [TTLMonitor] ERROR: error processing ttl for db: admin 10065 invalid parameter: expected an object ()
[23:52:07] <tomlikestorock> I'm trying to add a new box to my replset, and I keep also seeing these in the log: [TTLMonitor] assertion 13436 not master or secondary; cannot currently read from this replSet member ns:admin.system.indexes query:{ expireAfterSeconds: { $exists: true } }