Log file Viewer
#mongodb logs for Wednesday the 11th of July, 2012
[02:30:46] <kuhrt> db.films.update({title: "Montreal Stories 1944"},{$unset:{details:{"Release date":1}}})
[02:31:20] <kuhrt> I realize ^ is wrong because it removes the 'details' document, which I do not want, I just want to remove a field from it
[02:47:30] <mids> kuhrt: db.films.update({title: "Montreal Stories 1944"},{$unset:{"details.Release date":1}})
[03:15:09] <aws_2012> I apologize if this has been asked before/recently, but has anyone had difficulty getting and keeping your secondary node sync'ed?
[03:22:19] <aws_2012> repeat question: I'm having trouble getting and keeping my secondary in sync with my primary, despite turning off writes to the primary. Can anyone advise what I might be doing wrong?
[03:50:37] <ashepard> trying again, hoping i might find a nugget of wisdom: having trouble getting and keeping a secondary's optime up-to-date, despite turning off writes. anyone have any ideas to try?
[07:21:13] <DigitalKiwi> for a channel of >300 names there sure isn't a lot of activity here, ever >.>
[07:21:35] <DigitalKiwi> don't say nobody needs help cause i've seen lots of (unanswered) questions ;p
[07:23:03] <mids> DigitalKiwi: yeah, though I also see that in some other established opensource irc channels
[07:28:03] <akaIDIOT> just joined #cassandra, #mongodb and #hbase yesterday, story's the same for all :)
[07:34:55] <ranman> if you need mongodb support use the mailing list, it's great, or JIRA, or stackoverflow
[07:36:15] <mids> although just describing the problem out loud often helps... rubber duck debugging
[07:49:33] <dstorrs> when using the '$or' operating for querying, is clause evaluation order guaranteed to be left-to-right?
[07:56:00] <DigitalKiwi> how can I tell what indexes I should add to make a query work better? just by experience or trying different ones or is there some tool that could help?
[07:58:22] <NodeX> general rule of thumb is this : if you have large data and are querying on a key then it should be indexed
[08:00:17] <DigitalKiwi> I get it for single indexes but if I needed a multi index I think I would be very lost :(
[08:04:11] <DigitalKiwi> also, I know having indexes slows down inserts, but how *much* does it slow it down? Like could it be enough that it's faster to drop the indexes and recreate? all the datasets i've had only take a few seconds to add an index, but getting the dataset updated takes a lot longer but i don't know if it's just how it is or the indexes...
[08:16:50] <dstorrs> and how are you doing the update? via update() or findAndModify()? (the latter is ~100x slower than the former)
[08:17:40] <DigitalKiwi> well update was the wrong word, i meant "updating the database to have data" the data is actually new
[08:19:59] <dstorrs> er...except, it isn't supported by the shell, just (most) drivers. so I braino'd the syntax above
[08:20:47] <dstorrs> e.g., in Perl it's $db->coll->batch_insert(). in Python ISTR the standard insert() can take multiple docs.
[08:22:43] <dstorrs> actually, I lie: I can help. http://stackoverflow.com/questions/6153614/mongo-save-save-more-than-1-records-at-ones
[11:54:12] <Bartzy> What is the size of the buffer of documents sent with each next() call to a cursor ?
[12:02:35] <Derick> The PHP driver does it like this: http://derickrethans.nl/cursors-in-mongodb.html
[12:10:43] <Bartzy> so when the client asks for more documents, they are served as fast as possible ?
[12:11:04] <Bartzy> It's possible that after the query is 'done' (first next() happened), the next next()'s would be slow?
[12:13:11] <Bartzy> but how the server knows what documents to provide next, if the query happened a few seconds ago
[12:14:38] <Bartzy> But the pointers are to the actual location on disk (or RAM, if the OS decides it should be there) of the documents, and not some pointer to the _id or something, right ?
[12:15:30] <Derick> for the mongo server RAM=disk, there is no distinction. And yes, they're pointers to the location of documents
[12:17:20] <Derick> Bartzy: http://www.mongodb.org/display/DOCS/How+to+do+Snapshotted+Queries+in+the+Mongo+Database has a good explanation
[12:17:54] <Bartzy> I have one more question about .explain(). When executing explain on heavy queries, it doesn't really end, and you need to wait a lot of time
[12:18:12] <Bartzy> This defeats the purpose of seeing what the db is doing in order to optimize it..
[12:23:51] <Bartzy> But then what's the purpose of it for long-running queries? Is it possible just to get the decisions it's going to make like index use and stuff like that - like MySQL's EXPLAIN ?
[12:34:42] <multiHYP> please someone answer that, I asked the other day and still haven't found how to do it. although I have been doing some other things ever since too.
[12:36:05] <remonvv> Anyone a rough estimate on how long it typically takes a repset to vote in a new primary after it detects issues?
[12:36:10] <multiHYP> how do I hash the input field's password to be able to compare it against the hashed one in the db?
[12:36:58] <deoxxa> i would suggest starting at https://www.google.com/search?q=mongodb+password+hash
[13:20:54] <unclezoot> hi - i see it's possible to see what points are within a polygon, but can mongo tell you which poly contains a specific lat/lon?
[13:24:09] <unclezoot> i can find the nearest 3 districts using the geo index and a centre point within each of the districts to approximate the centers
[13:37:43] <fredix> I have two persistante connection to mongodb, but when I insert data from the one, sometimes I can't get the data from the other connection
[13:39:52] <mids> fredix: http://www.mongodb.org/display/DOCS/getLastError+Command#getLastErrorCommand-WhentoUse
[13:42:32] <mw44118> How do I say "give me the documents that have a description field that has at least one letter in it" ?
[13:49:24] <mw44118> NodeX: checking if the field exists doesn't seem to do what I mean; the field can exist, but point to a NULL value (or whatever you guys use for NULL).
[13:50:58] <mw44118> NodeX: I wasn't sure if I just wasn't aware of how to do it vs it could not be done
[13:51:04] <Derick> If you want to check for character count, you need to store that in the document
[14:03:14] <algernon> if you want post-processing, or long availability of logs, then indeed, capped collections are not good
[14:10:49] <jiffe98> trying to do so I get the error "errmsg" : "please create an index over the sharding key before sharding."
[14:29:22] <Bartzy> algernon: Why not just use regular ol' files ? If you're not adding indexes on the capped collection, it's pretty much the same
[14:29:52] <Bartzy> algernon: And that's only for application logs? errors and stuff? I didn't understand why many examples are for using capped collections for user interactions. Something needs to process those user interactions.
[14:30:34] <algernon> Bartzy: because I want structure. I can do that with files, but with mongodb, it's a lot easier.
[14:30:58] <algernon> Bartzy: nothing prevents you to send logs to a capped collection AND a non-capped one.
[14:31:35] <Bartzy> I just don't understand what's its purpose other than a fixed-size bucket for logs
[14:32:31] <algernon> plus, you can put a erm... what's it called.. well, a cursor on the capped collection that will follow the data coming in.
[14:33:02] <algernon> so you push logs into a capped collection, and you have your processing application listen via the cursor, do whatever processing you need, and send the result off wherever you want.
[14:33:28] <Bartzy> What happens if the processing application dies? The cursor dies with it, and some logs are lost (to processing, at least) ?
[14:36:06] <NodeX> [15:33:01] <algernon> plus, you can put a erm... what's it called <-- tailable cursor ?
[14:36:06] <Bartzy> and this data - is in its own capped collection, "spent_logged_in", or some general logging capped collection ?
[14:42:06] <algernon> http://mojology.madhouse-project.org/log/4d28cc53f310ef4f00000020 <- I like to do this kind of stuff.
[14:42:54] <algernon> combined it with an external indexer (solr, lucene, ES, whatever) and it's wonderful.
[14:45:00] <algernon> one big logs collection, a few stats collections and a capped recent collection
[14:45:26] <algernon> I'm not entirely happy with the way I deal with stats, but had no time to come up with anything better yet.
[15:00:05] <timing> Hi! I have a collection with a few basic fields and one field contains a large multi dimension array. I don't mean to query on the data in the array. So can I just put the array in the collection as well. Or is it (performance wise) better to just serialize the array myself as string before insertion? So it just becomes a large string instead of a 'real' array
[15:01:41] <timing> So the data can be something like this: {field: 'value', bigArray: '[[1,2,3,4],[603,123,2039]]'}
[15:47:01] <Bartzy> why Mongo saves numbers as double by default? Isn't that very wasteful in space ?
[16:28:57] <Lujeni> Hi - i need move many documents between two servers. Should i use mongodb command as mongodump / mongorestore ? or store all document in variable and use "forEach" to insert?
[16:30:39] <SLNP> We are seeing a strange issue at the moment, we are hitting a connection limit of 1000 on our mongo rep set and both mongo instances are crashing
[17:31:15] <timing> is it recommended to serialize it first, or shall I just 'put it in' the collection?
[17:32:10] <timoxley> timing if you don't need to query on that data, then perhaps serialising it is a good option
[17:35:09] <timoxley> timing don't really know, just seems to make sense for you to put things into the db in the most primitive form you need them as
[18:02:33] <redalpha> If i have 2 servers setup with replication - and both servers have an application connected to their own mongodb - and both applications doing writes and reads, will that go well ?
[18:11:20] <armandop> lets say im creating a mobile app and the data we collect is just for a specific user, comments, and other users do not see these comments uts just the user
[18:11:59] <armandop> i want to create a single collection called users which contains a comments property which holds an array of Comment objects.
[18:15:28] <mollerstrand> armandop: a collection is created as soon as you insert a document into it.
[18:16:03] <armandop> let me rephrase it. should I, given the conditions stated, create two collections?
[18:16:26] <armandop> a users collection and a comments collection? and use a DBRef to link both. I think this is totally the wrong approach to take
[18:16:44] <armandop> since the comments are not shared and can be easily be embeded into the users collection.
[18:18:13] <mollerstrand> to me it sounds valid to have two collections, one holding user documents and one holding comments documents
[18:18:55] <redalpha> i dont like to nest a lot of diffrent entities into 1 collection - i tend to create collections per entity types
[18:22:14] <armandop> thats what i used to formulate my assumption on architecture behind my reasoning.
[18:29:20] <redalpha> its just my preference to do manual referencing rather then embedding and DBrefs
[18:30:30] <redalpha> maybe if you do a lot of reads - where the referencing would mean doubling the amount of queries u execute
[20:37:02] <timing> when I have a comment attached to a post i can give the comment a post_id. what can I ise best the the id? the MongoId object or just the numeric $id value?
[20:44:20] <bluejade> where is the best place to ask beginner / intermediate questions related to structuring and querying mongodb documents? here? mongodb-user google group?
[21:23:25] <mrfloyd> Any idea how can i collect a nested object from a document, and only this object not the whole document
[21:34:07] <armandop> @mrfloyd - have yu tried db.collection.find({your where clause}, {property_you_want:1})
[21:43:48] <armandop> alex is the primary key (if your looking at this from a Relational DB mindset) using that key you associate a post to alex
[21:44:18] <armandop> you could use the internal object id that is created for each document but it gets dicey when fetching info using code.
[21:45:48] <armandop> {_id: 697698779886hhy76 } <-- starting object example. when you call it from code if you convert the object id to string it will no translate to the object id. object id != (string)object_id
[21:46:39] <domo> ah i see, but my driver in the app allows me to create MongoIds as objects from strings
[22:00:07] <hadees> is there anyway to see what queries are being run? I have a mongodb query that works but when I converted it to php it doesn't. So now I want to see what is actually being run.