Log file Viewer
#mongodb logs for Wednesday the 16th of January, 2013
[01:59:23] <aZnmAn> Hi guys -- new here. Wondering how to find() a nested document... http://pastebin.com/z9Zd9j6g for JSON. Need to find posts by an author (I'd like a nested setup for categories too)...
[05:37:18] <domo1> whats the difference between using mongo references vs referencing doc IDs manually?
[06:04:50] <kreedy> anyone have a good first place to look for why mms only sees one secondary and no primary in my 3 box replicaset? when i added all 3 boxes with auth. there is also an agent running on each one
[06:08:54] <jeremy-> Is there anyway to use update/upsert to NOT do the upsert (create an additional entry), of a second query is true. I can think of one way of doing it which is doing a .find() first, then just programatically not doing the upsert query if that is true. Anyway in one mongo request? I suppose its wanting an additional query parameter for upsert that i dont see.
[06:11:13] <jeremy-> my question was kind of vague so I'll give an example from my pymongo: self.collection.update({'date' : q}, { '$set' : {type : q, typestatus : 'ready'}}, True) IMAGINE THERE IS AN ADDITIONAL FLAG... DONT DO UPSERT/ANYTHING IF... typestatus : { '$nin' : ["csv", "complete"]} ..
[06:26:22] <abhi9> is there any way to set matched item as a key from an array in mapreduce, except loop?
[08:25:56] <ranksu> the thing is that I'm having a large collection of POI's in mongo. When I'm doing a bounding box query, I got the POI's only in center of the box. however I'd like to have them in newest first
[08:26:43] <ranksu> that's the problem since it doesn't seem to work. But stack is java with spring data
[08:34:34] <ranksu> I do have a matching index. the query _seems_ to run fine in console, but from spring it's slow. I've tried to debug the query, but I really cannot see the proper query that's been sent to mongo
[08:43:14] <kali> well, at least try to check out what spring does: you can probably bump the log level in your java app, or use mongosniff to sneak on the network, or use the mongodb profiler to get all queries
[09:21:54] <waheedi> i just check my servers and saw oplog configured as oplog size: 9118.617382812501MB / log length start to end: 8378451secs (2327.35hrs)
[09:54:58] <ranksu> one more thing. in mongo shell, is it possible to use a unixlike more/less since not everything fits in my terminal window?
[10:21:16] <ron> chickamade: now... try asking the same question again, only this time make a little more sense ;)
[10:23:33] <chickamade> can you ask mongodb to spit out the query plan without actually running the query?
[10:25:48] <Baribal> Hi. A friend of mine just tried to use a collections .copyTo (without having found API docs on it, btw), using a collection as parameter. The result was: "SyntaxError: missing exponent (shell):1". So... Does .copyTo() actually copy collections? And more importantly, what does copyTo do and where's the docs?
[10:27:30] <ron> chickamade: it's not a bad suggestion though. maybe you should open a feature request.
[10:58:14] <xcat> My application would find it easier to just ensure the index exists than to try to determine if it exists first
[10:58:35] <xcat> In other words, it would be easier to blindly call ensureIndex than to try and introspect the collection to figure out whether the same index already exists
[11:00:27] <Derick> xcat: most drivers cache the ensureIndex, so you can call it as much as you want
[11:01:33] <xcat> I want to do this with the TTL index; it won't have any adverse effect there either? Like restarting the background theread or anything weird
[11:01:41] <NodeX> what happens if you've got a massive collection that has no index and your app creates one...... it will hang your app
[11:03:39] <NodeX> xcat : it's exactly what you're talking about .. you asked about creating indexes over and over in your app, and i explained why it's a bad idea
[11:54:04] <ABK> issue is we currently don't wanna change in Prod., so to have a similar version across require it
[11:54:43] <ABK> we'll only switch in Prod after regressive and long testing... that will take time... so for until then we are stuck
[11:55:05] <Derick> http://downloads-distro.mongodb.org/repo/redhat/os/x86_64/RPMS/ is all we have online
[11:57:08] <ABK> I understand the reason for not using that... but as I said pushing the infra change to Production wouldn't be without enough testing
[11:58:30] <Derick> also, what you build from source is not going to be the same as the RPM I suppose
[12:00:57] <ABK> yeah the spec seems to be there @ https://github.com/mongodb/mongo/blob/master/rpm/mongo.spec
[12:01:21] <Derick> do note, that we probably not support this though and really recommend upgrading to at least the latest 2.0.x
[12:02:03] <ABK> sure... this is just for the new upcoming nodes using same config-management-tool with fixed Mongo version due to Production
[13:02:11] <lgbr> my reduce function only seems to be called on a very limited set of my database. So I might have hundreds of objects in a collection whose map key (first parameter of emit()) should have hundreds of entries in it, but when reduce is called, only ~70 or so items are in the second parameter of the reduce() function. Am I understanding map reduce wrong?
[13:24:30] <kali> lgbr: the reduce is only called is there is something to reduce: if you only have one value emitted for a given key, it will not need a reduce
[13:25:11] <kali> lgbr: also, consider using the aggregation framework instead of m/r if it can handle your task as it is much more efficient
[13:26:11] <lgbr> kali: My problem is that I have a key that should have >100 values, but it is consistently calling reduce with <80 values. They're not being broken up into different reduce calls or something are they?
[13:27:06] <kali> lgbr: if a key have 1000 emitted value, mongodb can choose to reduce the first half, then the second one, and the reduce again the two values
[13:28:47] <lgbr> kali: Hmm. The second parameter of my emit() function should have the same format as the return value of my reduce() function?
[13:36:49] <jonesy> I use Diamond to monitor mongodb, and it's creating metrics for, like, every mapreduce job. Where can I learn more about what's happening behind the scenes when I launch a mapreduce job, so I can better understand how to work with my monitor's data collector?
[15:52:56] <rickibalboa> Can anyone help me out with an issue. I have a query which I can run directly on the database https://gist.github.com/4548097 and it returns results which is fine, the next query is exactly the same, just PHP-ified, I run that in my PHP code and I don't get any results, database is definitely connected etc. Any ideas?
[16:36:02] <owen1> what triggers an election for a new primary? only loss of connection to a member or are there more conditions like high cpu/memory/diskspace?
[16:44:46] <Derick> StarX: because the algorithm to vote the primary doesn't scale beyond that (I think)™
[16:47:35] <StarX> Derick, hmmm.... for my case... i need 5 vote members and 20 non-vote members.....12 server limit is kill me :/
[16:54:26] <StarX> 5 servers in Replica mode.... and 20 "slaves" (from Master/Slave mode) using (if crash) the 5 "masters" in replica
[16:56:03] <StarX> kali, i need all local data.. in 25 servers... using Replica i limited to 12 servers
[17:01:01] <NodeX> I think he means he want's 25 copies of his data in 25 servers so he can read from them
[17:06:12] <StarX> NodeX, my ideia is use mongodb... for know how many time... each "email address" receive a "email" for each clients.
[17:17:10] <WarheadsSE> True, but assuming I'm here beacuse I want my Android phone to run a nodejs webchat server with mongodb does't help in understanding the problem in any way either.
[17:17:54] <WarheadsSE> im not worried about that, I am here specifically about mongodb & it's ASM opcode failures.
[17:21:45] <WarheadsSE> Here is what we are hitting on a compile for armv7-a vfpv3-d16 http://sprunge.us/QeOT
[17:23:22] <WarheadsSE> build scripts are located here: https://projects.archlinux.org/svntogit/community.git/tree/trunk?h=packages/mongodb
[17:34:31] <owen1> i got 2 datacenters. is it fine to have RS of 3 nodes, 2 on east coast and 1 on west?
[17:39:09] <owen1> the only issue is if i lose 2 primaries, i'll have to manually set the last one to primary
[17:40:22] <owen1> so if i'll have total of 5 i can lose 2 primaries and a 3rd one will become primary.
[17:49:06] <evan_> I'm doing some very detailed reporting on 100million events and was looking to use map-reduce. Should I use cassandra or mongodb or what?
[17:58:47] <WarheadsSE> Saturation is the user that has me looking for a way to get the package fixed in our builds.
[18:58:20] <owen1> i don't understand the requirement for odd number of nodes. let's say i have 4. primary dies. 3 can vote. what's the problem?
[19:03:09] <strnadj> You need ODD number of voters, because if you have even number of votes, there can be a situation - no one of nodes has most votes - 4 = can 2:2, 6 = can be 3:3 etc.. if you have ODD number of voters - this situation is not possible 7 - can be 3:4 etc...
[19:05:41] <owen1> JaredMiller: if it's the secondary, nothing happens. if it's the primary, a new primary will be elected
[19:08:58] <kali> the problem is not tie during the votes process. the problem is to make sure there is only one election taking place at a given time. so for an election to occur, you need a strict majority of nodes to be able to talk to each other
[19:23:19] <tworkin> in pymongo, there is collection.find().count(), but why isnt there len(collection.find()) ?
[19:29:05] <kali> owen1: http://docs.mongodb.org/manual/tutorial/reconfigure-replica-set-with-unavailable-members/
[19:53:39] <owen1> i had 1 primary that survive out of 3. i did some config change to one of the secondaries (added himself and the other secondary to the conf) and reconfigure it. it seems like the survivour primary is not aware of the 2 nodes that came back to life.
[19:55:00] <owen1> i thought the configuration file will be shared whenever i change it on any host.
[20:35:51] <WarheadsSE> kali: NodeX would either of you know who might be able to lend a hand in sorting out the issues seen by my users hwne building mongodb?
[20:37:39] <kali> WarheadsSE: you need help from the actual developpers... most of the people are are users, and most of discussion here is about usage
[21:17:19] <andrewwinterman> do you guys know when you'd want to use multiple databases rather than multiple collections?
[23:02:26] <owen1> let's say i have 3 nodes (primary is here too) on west coast and 2 on the east. if the connection between the coast is lost, what ensures that i don't end up with 2 primaries? is it because of the fact that the east coast is not a majority?
[23:10:06] <owen1> another question - is there a need to restart my node client after reconfiguring a replica set (let's say new primary is elected, etc)?
[23:12:39] <linsys> owen1: What client are you running? I think node.js might have a bug if auth is enabled but besides that no to your restarting of the app node question.
[23:13:29] <linsys> owen1: for a member to become master it must have majority votes. If the west coast went down, 2 nodes is less the majority so no nodes on the east would ever become master
[23:14:28] <owen1> linsys: i use the 10gen official client. not sure if auth is enabled. what does enabling it gives me?
[23:14:54] <linsys> owne1: Well there are 10gen clients for many things.. what programing language is your app in?
[23:18:17] <linsys> owen1: it means you would need to pass a username/password when your app connects to mongodb, same when you connect with the shell
[23:18:32] <linsys> vs now where you can just do like mongodb --host <ip> and look at all the dbs and such
[23:20:34] <owen1> linsys: regarding the other question, what if i have 1 host(primary) and the west and 2 secondaries on the east and there is a network issue between east and west. the 2 secondaries might think that the west is down so they will elect a primary and i'll have 2 primaries.
[23:23:06] <linsys> No, if that where to happen if there is a network issue the primary in the west will demote itself since it lost the majority of the vote thinking it's isolated
[23:26:41] <owen1> so if the primary is among minority, it will demote itself? let's say i have 2 nodes(1 of them is primary) in the west and 3 nodes in the east, and there is a network issue, the primary will demote itself as well?