Log file Viewer
#mongodb logs for Tuesday the 29th of September, 2015
[00:06:13] <fedora_newb> joannac, maybe I can get your help on one more item here, http://screencast.com/t/QmemJ5GBN
[02:45:38] <fedora_newb> Parameter 1 to MongoCollection::insert() expected to be a reference, value given. I am not finding any sort of clean solution for this anywhere. Anyone have an idea on this one?
[03:46:59] <nahtnam> But on my site, I'll be parsing protobuf files of matches that contains a list of kills, deaths, wins, losses, and it seems like mongodb would excel at storing match info...
[07:20:02] <bjpenn> anyone use mongodb with ubuntu? i installed it via instructions here, and the `service mongod start` script doesn't seem to work
[07:20:11] <bjpenn> anyone use mongodb with ubuntu? i installed it via instructions here, and the `service mongod start` script doesn't seem to work
[07:36:47] <zamnuts> brucelee, ever? it could've been very fast, can you check the logs? should be in /var/log/mongodb by default
[07:43:00] <zamnuts> i'm probably the only person who doesn't run anything in AWS, so i have no idea how AWS exposes the filesystem
[07:44:28] <jokke> is a query like this possible? { 'embedded_docs.name': 'foobar', 'embedded_docs.$.value': { $exists: true } }
[07:44:29] <zamnuts> brucelee, mongod will run as user "mongod" ... does this user have permissions on /data1?
[07:44:56] <jokke> so that the $ operator would be filled with the index that match the first part of the query
[07:50:19] <zamnuts> mongod.lock exists, but is owned by root:root, which is why user mongodb cannot write to it :)
[07:51:31] <zamnuts> ...at least i HOPE that's the case, not sure why root would own the lock file; i only administer mongodb on RHEL, so....
[08:01:28] <zamnuts> brucelee, that warning is an issue with the specific hardware/OS that node is running on
[08:02:30] <zamnuts> the readahead value says how much EXTRA data should be read on any given read; this is beneficial because sequential operations are always faster than random ops
[08:03:43] <zamnuts> a large readahead is great when there are large documents that extend beyond the block size, i.e. where sequential reads are helpful; but a large readahead is wasteful on small random I/O
[08:06:14] <zamnuts> the warning in the log is saying that 2megs as a readahead is too large, and that the recommended size is 256K (0.25megs)
[08:07:38] <zamnuts> brucelee, how-to for ubuntu @ http://askubuntu.com/questions/500491/setting-persistent-readahead-buffer-other-than-in-rc-local
[08:14:23] <brucelee> lol zamnuts i was reading up on mongodb stuff on google and came across this interesting link
[08:41:39] <ahihi> hi, I want to update an array in a document in a manner that depends on its current state - if the array's last element matches a certain condition, I want to update that element, otherwise push a new one. is this possible to do without two round-trips to the DB?
[09:33:51] <pamp> I need to create a user with permissions to read any db, and can create their own collections and be admin of their collections
[09:35:14] <pamp> the first role is readAnyDatabase, but what is the role in order to create and manage their own collections?
[11:46:28] <pagios> http://cryto.net/~joepie91/blog/2015/07/19/why-you-should-never-ever-ever-use-mongodb/
[11:51:52] <Zelest> They're absolutely right about "never use a database because 'everybody else does', do your own research as to the advantages and drawbacks of a particular database." though..
[11:52:07] <Zelest> I guess that's what they did.. without doing it right.. and apparantly they weren't happy with the results.. *shrugs*
[11:55:58] <Zelest> I love Postgres as much as MongoDB but for various things I find one of them better than the other.
[11:56:36] <Zelest> For example, the automatica failover feature in MongoDB's replicasets is lovely for high-availability and insanely simple to setup..
[11:57:05] <Zelest> When it comes to fulltext search, I prefer PostgreSQLs tsearch2.. Lookup your needs and decide there after.
[12:24:27] <fedora_newb> Curious if anyone here uses laravel and mongo together that may be able to help with this error, Parameter 1 to MongoCollection::insert() expected to be a reference, value given
[12:30:24] <sedavand> is it possible to do a .find() query that matches against a list in another colletion, without having to write a JS function?
[13:06:25] <Rumbles> Hi, one of our devs has asked me to set up an arbiter on 2 nodes (of 3) in our dev environment mongodb cluster, I have checked the docs and they say not to do this. He has said he has done this on 2.6 and had no troubles. Now I don't want to make a load of work for myself and I want to find out what I could break if I do this, can anyone advise why it is advised not to set up an arbiter on a live mongodb node? Or am I no
[13:06:25] <Rumbles> t understanding something? I'm pretty new to mongo (but the dev has been using it for years)
[13:06:58] <Derick> Rumbles: so, out of the three nodes, 2 will be an arbiter and 1 a data carrying node?
[13:10:39] <Derick> 1. there is no point in doing it. The idea of an abiter is to be able to vote if there is not a majority. Putting an arbiter on two of your three live nodes does not help
[13:11:57] <Derick> no matter which physical node would go down, there need to be two nodes still up to vote... and you already have 3 nodes so that works just fine (with one node down). You can't reach a majority of *two* physical nodes go down even with the arbiters on them
[13:12:56] <Derick> secondly, you get two mongo processes fighting for resources. That's not so bad with an arbiter + real node, but it's not helpful.
[13:30:43] <Rumbles> Derick, one of the devs (well our CTO) just went ahead and did it, and I got the infrastructure slightly wrong, we have 3 nodes, and 3 arbiters, they wanted one arbiter setting up on an active member...
[14:30:24] <Rumbles> thanks Derick, like I said earlier, I didn't set it up, I get why it's pretty much pointless.... I'm guessing the advised way to use arbiters is if you have an even number of mongo nodes use 1 arbiter, otherwise if you have an odd number, you don't need any... ?
[14:31:02] <valera> from what I understood mongo is not checking utf-8 correctness on document insert, however map-reduce engine fails with exceptions if it spots broken utf-8 character sequence, is there any way to figure out what document caused an exception ?
[15:01:27] <deathanchor> having a pymongo2.6 issue using the same query on mongo works, but not when I run it through pymongo find
[15:03:16] <deathanchor> so find({ array : { "one" : "yes", "two": "no" } }) works on mongo shell but not when I use pymongo find
[15:09:09] <GothAlice> If PyMongo doesn't like OrderedDict (it should…) you can always drop down and construct the ordered BSON yourself, too. :)
[15:24:45] <deathanchor> GothAlice: that fixed it all, didn't realize that order mattered to mongo when doing a sub doc comparison.
[15:25:19] <GothAlice> MongoDB does binary (hashed) comparison of whole sub-documents. Changing the order would change the hash. :)
[15:58:26] <Snicers-work> I need find documents in a collection that have matching fields. For example all users with duplicate emails, without knowing the values of emails.
[15:59:39] <StephenLynx> after you finish grouping, use a $match to filter only the documents where the incremented field is > 1
[16:01:57] <Snicers-work> StephenLynx, what do you mean use the email as the _id, the _id is already set.
[16:14:51] <Snicers-work> currently I have this but it doesn't show which are duplicates: https://gist.github.com/anonymous/fab9673b7c158e39b443
[18:59:27] <devdvd> Hi all, quick question. Did the way capped collections work between 2.4.5 and 3.0.6 change? the reason i ask is because im trying to migrate from 2.4.5 to 3.0.6 and did a mongodump on 2.4.5, when i go to import into 3.0.6, i get this message "error creating collection searchCountsByMember: error running create command: exception : specify size:n when capped is true. Here is my searchCountsByMember.metadata.json {"options":{"create"
[18:59:27] <devdvd> :"searchCountsByMember","capped":{"$undefined":true},"size":{"$undefined":true}},"indexes":[{"v":1,"key":{"_id":1},"ns":"prodege.searchCountsByMember","name":"_id_"}]}
[18:59:27] <devdvd> . Now I see the problem and I thinki know how to fix it (i fixed it on another one like this by defining a size). Is that the proper way to go about it or am i missing something?
[20:02:14] <pokEarl> So I am a bit confused about the ObjectID vs just having the actual identifying string of the object id. Since the object id itself has more info hidden in it than just the string. If you send over id string of the object id through something like a database layer that sends Json, and then create a new object id on the other side with the same string identifier, does it matter?
[20:18:17] <arasmussen> anybody have an idea about http://stackoverflow.com/questions/32852646/find-and-update-multiple-documents-atomically ?
[20:18:42] <idd2d> I'm modeling a stock tip with MongoDB, and one of the things I need to track is the stock's value over time. Each stock tip will have a bunch of stock value snapshots taken every 10 minutes or so. I'm wondering whether I should associate these values via a relational ID, or simply embed the value snapshots in the stock tip schema.
[20:19:21] <idd2d> One requirement: I need to be able to query for a specific slice of the stock value snapshots. i.e. I want to see how this stock value fluctuated between Sept. 29 and October 2.
[20:19:42] <arasmussen> idd2d won't you have a huge number of these snapshots then? should probably each be their own document, no?
[20:20:46] <idd2d> We're still trying to work out just how often we'll be grabbing data, but it will be on the order of every 10 minutes for any given stock tip, for around a week, on average. That's about 1000 data points per stock tip.
[20:22:09] <idd2d> My initial thought is that, since each stock tip has just one set of value data, it makes sense to simply embed it. I see two issues, potentially: That's a lot of embedded data, and I don't know if I'll be able to effectively query for a time range within that data.
[20:23:14] <idd2d> Also, is it possible to query for a stock tip *without* the embedded value data? That way we're not pulling all these data points on every request
[20:23:16] <arasmussen> yeah I would probably lean towards having a separate collection of snapshot data
[20:24:16] <arasmussen> yes you could unselect it with db.collection.find(query, {snapshotData: false})
[20:24:51] <arasmussen> but I'd still lean towards having a separate collection of snapshot data with {stockID, timestamp, value}
[20:25:25] <idd2d> The downside of such an implementation is the extra work of associating each datapoint with its stock tip
[20:26:48] <idd2d> Right, but when selecting, Mongo is having to do extra work behind the scenes to associate any data point with its tip, right?
[20:27:44] <arasmussen> Then you could easily query db.snapshots.find({stockID: 123, timestamp: {$ge: ONE_WEEK_AGO}}) kinda thing to get a list of snapshot datas from the past week
[20:28:24] <idd2d> haha, gotcha. My questions are based on only a cursory knowledge of the nosql world
[20:32:38] <idd2d> This is a pretty good representation of my (limited) knowledge of mongo: http://blog.mongodb.org/post/87200945828/6-rules-of-thumb-for-mongodb-schema-design-part-1
[20:34:13] <idd2d> Where A has_many B, give A a list of ObjectID's representing its constituents (best used in one-to-many relationships)
[20:34:45] <arasmussen> Anybody know if there's a way to effectively figure out which documents you've modified? Or atomically find and update multiple documents?
[20:35:04] <idd2d> And where A has_many_many B, put a reference to A in B (best in "one-to-squillions", as the author calls it)
[20:35:40] <idd2d> It seems clear to me that embedded documents aren't a great idea, because there's just too many data points. So what about the other two options?
[20:36:54] <arasmussen> idd2d: yeah you'd want the last one where each of the snapshots has a reference to the stock it's associated with (A has_man_many B)
[20:42:37] <idd2d> Gotcha. Is there any benefit to using the "one-to-many" scenario specified in that article?
[20:48:32] <arasmussen> the convenience would be you'd be able to fetch all the data in a single document/query
[20:49:35] <arasmussen> but with two separate collections you'd be able to fetch everything in two separate queries, I don't see anything wrong with that
[20:50:35] <arasmussen> it might take up a bit more space but the whole thing is simpler, you don't have to unselect certain data, you don't have enormous documents, it's straightforward to do to queries like "timestamp > one_week_ago", and still super performant as long as you have indexes
[23:31:37] <blizzow> I have a database foo, with collection bar, with documents baz and bum. The baz and bum documents have two string fields, _id and queuedblock. I'm trying to do an eval statement from the CLI to return the value of queued block but it's not working. I can't even get a list of documents from this: mongo -u user -p password mongohost/foo --eval "db.bar.find( {} )"