Log file Viewer
#mongodb logs for Monday the 17th of September, 2012
[00:30:19] <timeturner> vsmatck: the thing is that I have a bunch of posts, which have one tag each, in a blog
[00:31:11] <timeturner> so I don't know if it would be safe to store the post objectids with the tag in an embedded arrray in the main blog document
[01:32:20] <igotux> say, i have 3 hosts for setting up replica sets and if the primary goes down, secondary gets promoted as primary.. but do i ensure my application will talk to thew _new_ primary?
[01:42:20] <eydaimon> is it possible to go backwards from the objectid to get a timestamp? IIRC the objectid is related to time somehow
[02:00:58] <igotux> if the client talks to mongod, then i need to update dns cname and point it to secondary manually..correct
[02:36:06] <Ventrius> I haz question onf replication/sharding: Is it possible to have 2 local servers that have their own data
[04:39:36] <igotux> say, my db will grow upto 500G of data .. hence am planning to use 3 servers , in which 2 servers are having 750G of disks and 16G of ram and one small node for arbiter
[04:43:32] <igotux> timeturner: as you said earlier, mongos available only if i use sharding ..correct ?
[04:56:14] <abstrax> i am storing created_at agaianst each record, do you recommend MongoDate or a timestamp integer?
[04:56:39] <abstrax> IN mysql, i would do a date because i get access to amazing date/time functions, but what about mongo ? need some suggestions :)
[05:03:04] <mrkurt> and unless you have a really weird workload, 500GB of data with only 16GB of RAM might be a disaster
[05:05:39] <igotux> you see any smokes in this design .. because for me, sharding doesn't look simple... and needs more nodes as well..so wondering any issues if i go with just Replica Sets only
[05:08:11] <igotux> when i should go for shards.. usually what should be your data size when people consider using sharding in production environment ?
[05:08:24] <chovy> my save seemed to work in nodejs, but when I type 'show collections' in mongo console, nothing is returned.
[05:12:49] <ron> chovy: you may have the client configured to ignore errors. I don't know node.js so I can't give you exact pointers.
[05:15:52] <chovy> i'm using mongoose. All the examples I've seen just pass an error callback to foo.save
[05:18:49] <timeturner> how many objectids do you think can be stored in an embedded array before it hits the size limit
[05:25:13] <mrkurt> if data is actually growing quickly, though, I'd probably start thinking about it at 300GB or so just to keep things manageable
[05:25:51] <mrkurt> you don't want to get too far in a hole, which means if you're ever going to have to shard you can't wait too long
[05:28:34] <mrkurt> all of this assumes you only do 10-1 ram, if you buy nice servers with >200GB of RAM you can hold off longer
[06:30:55] <igotux> so if i use only replica sets, if the primary goes down, i need to manually ask the application to connect to the new primary..correct ?
[06:33:21] <timeturner> which is why you can basically type "mongo" in bash and start interfacing with mongodb
[06:38:30] <igotux> timeturner: so in my 3 node replica set, apart from starting mongod process on node-1 and node-2 , do i need to start mongos process as well ?
[06:40:33] <igotux> so in my use case, if the primary goes down, i need to point the application to secondary IP ( which is the current master) or change the cname in dns
[06:40:43] <timeturner> then you will need to manually fix whatever the problem is and rejoin it to the replset etc.
[06:42:33] <wereHamster> mongo will re-elect a new master, and the driver should automatically switch to it
[06:43:23] <igotux> wereHamster: my driver will be aware of only the master ip na ?.. or is it like, we need to feed both primary and secondary IPs to driver ?
[06:43:56] <wereHamster> igotux: you should give the driver more than one IP, but it will also work if you give it one
[06:44:18] <wereHamster> igotux: of course, if you give it one and that server is down while your application is trying to establish a connection to your replset, it will fail
[06:44:51] <timeturner> that way it can figure out who is the new master or who it should read from etc.
[06:44:57] <wereHamster> igotux: so you usually supply multiple addresses (does not need to be all) and the driver will discover all the nodes
[06:46:27] <wereHamster> igotux: have you even read http://www.mongodb.org/display/DOCS/Replica+Set+Tutorial ?
[06:47:24] <igotux> read the almost same content from other source.. not doc.mongodb.. will chk this out as well
[07:04:35] <timeturner> for example, there is a users collection in which each user has an array of objectids in which each is an objectid to a single "tag" document in the tags collection. The tag document itself doesn't have much value but the objectid of the tag document is referenced by many post documents in the posts collection. would it make sense if I were to query for all posts that have the tag field equal
[07:04:35] <timeturner> to any one of the ObjectIds stored in the user's field of an array of ObjectIds?
[12:14:34] <ezakimak> how do you create/save a server-side function from pymongo? I can do it from the shell, but the examples i found don't work from python: db.system_js.funcname = "function(parms) { ... }"
[12:18:37] <ezakimak> code being a string containing what? eg "function(<parmlist>) { <body> }", right?
[13:29:20] <brahman> Hi, has anybody seen what appears to be an issue with the connection pool having stale connections. We are using version 2.0.3 in a sharded environment (aiming to upgrade) and every now and again we are getting no documents being returned but we are certain the documents exist. after restarting mongos the issue resolves itself.
[14:11:08] <rydgel> Does cloneCollection() works with sharding and mongos. Can it be run inside when connected to a mongos? I've got weird assert errors.
[14:56:35] <Lujeni> Hi - the best order for shutdown properly an "shard" : mongoS , then mongoRS ( node , then arbiter ), then Config server ? thx
[14:56:37] <kanzie> Im trying to add a property to an existing collection using the inc-operator. If the field don't exist since before, will this be a problem?
[14:57:44] <kanzie> like I have users-collection with 100 users, but no one has the property age… now I want to add it if it doesn't exist (with value 1) or increment it with 1 if it does exist
[14:59:55] <NodeX> db.foo.update({foo:1234},{$in:{field:1}},false,true); will add "field" = 1 if not exists or increment it by 1 if it does
[15:34:57] <Truquero> suddelny mongodb is having problems i cant start it and when i try to connect to it through an application i get this error: Could not connect to a primary node for replica set <Moped::Cluster nodes=[<Moped::Node resolved_address="127.0.0.1:27017">]>
[15:54:57] <VinX> I would like to wrap mongo DB server in c++. I mean try to use it as "memory managment".. dunno if my will is clear.. :P
[15:55:30] <Venemo> could someone please tell me what's the mongo alternative of an SQL varbinary or image column?
[16:01:25] <Venemo> Vile1, so, I should use GridFS and my documents should reference a file from GridFS?
[16:10:34] <Venemo> my other question is, has anyone managed to get MongoDB working on ARM? I see there are some guys who claim to have done it, but I haven't found any code or guides on how to do it.
[16:27:53] <Venemo> algernon, well, I only intend to use this ARM device for testing, so I can live with that.
[16:28:31] <Venemo> algernon, the problem is that I didn't find any packages or build instructions. only a comment in a bug report from someone who claims he did it.
[16:29:15] <algernon> Venemo: if you're on armel, it shouldn't be *too* hard to compile it, I believe.
[16:37:05] <Venemo> I actually haven't tried. I just googled for it and the only thing found was an unresolved bug report. that's why I thought I'd ask here first :)
[17:03:43] <stefankoegl> I'm trying to build mongodb from git master, and get the following error: https://friendpaste.com/55MrWtpTIufnEMRrxXbNyJ what am I doing wrong?
[17:04:22] <jmar777> quick sanity check - does a TTL collection with a second index (besides the expiry value) for username/ip/other-identifier seem like a reasonable way to keep track of failed login attempts within a certain window?
[17:16:12] <jgornick> I see if I do, myObjectId.toSource(), it returns an object with a str property.
[17:19:57] <xamox> Is it okay to delete all the files in /var/lib/mongo? I want a clean slate, replication and all, is it okay to blow that folder out?
[17:24:02] <jgornick> NodeX: When I do String(some object id instance) in a map function, the typeof returns "string" and the value is a string as "ObjectId("the hex")"
[17:25:02] <jgornick> NodeX: https://groups.google.com/forum/?fromgroups=#!topic/mongodb-user/UhH3PO5VN6M
[17:28:52] <jgornick> NodeX: Yeah, I think the issue that was filed is wrong. They should fix the default String type conversion in 2.2 back to returning a string of the hex string.
[17:56:09] <wladston> when I check the mongo profiler, it doesn't show info about the subset of fields I passed to my query.
[17:56:44] <wladston> is this a problem with pymongo ? or the profiler doesn't care about the subset of queries that are requested ?
[17:57:08] <wladston> (I'm trying to optimize counting elements by only selecting the "_id" field ...
[17:59:51] <smartviking> I have a db with documents, I want to create additional functionality and need to update the old documents to hold a new value. What do I do?
[18:00:41] <smartviking> I've tried a dozen ways to add a 'popularity' field and they always fail, I feel as if some of them ought to have worked
[18:06:21] <doubletap> unless a single argument call implicitly is saying db.mycol.update({}, {$set:{the:'house on fire'}});
[18:09:24] <smartviking> Ok turns out I made a typo, and used a non-existent database, that explains a lot
[18:10:07] <wladston> I'm selecting a subset of fields in my query, but the mongo profiler doesn't show this information
[18:38:53] <cmendes0101> In the aggregation framework, im attempting to do a $group but is there a way to modify the data before the $sum? Like I'm trying to replicate a GROUP BY in SQL with it retrieving another field with a CEIL()
[18:42:35] <cmendes0101> $group : { _id: "$cust_id", count: {$sum: 1},minutes: { $sum: $length}} <- To do a CEIL on $length
[18:44:15] <Neptu> hej someone that used mongodb c++ driver when i use the .connect("localhost") do I need to add all the hosts of the replica set like I do with the python driver??
[18:58:27] <Neptu> I read some time when i was using the python driver that was not working so thats why im asking
[19:49:46] <BrianAnthony> hey guys. receiving the error "Can't take a write lock while out of disk space" when trying to insert a record into mongodb. the collection is brand new but it does have sharding enabled with a record inserted prior for creating the index and sharding on the appropriate keys. anything fishy here?
[19:54:11] <BrianAnthony> well lucky me i'm on the dev on the project so i get to trouble shoot without access to the box because i'm not on the networking team =p
[19:56:44] <BrianAnthony> well i've been googling and it's not like "oh it's just this" it's either some crazy edge case or we're out of space and i'm being told we have space...
[19:58:36] <ron> okay, if you need to support the database, you need access to the database machine. if you don't need to support the database, then tell whoever is supposed to do it to do it.
[19:59:13] <ron> what you describe now is the best way to make problem solving take exponentially longer.
[20:13:40] <BrianAnthony> so i am being assured 100% there is available space for folders specified in the dbpath.
[20:15:30] <BrianAnthony> one thing that stuck out to us from the logs was a warning saying "warning: could have auto split on collection: my_collection.collection_name but: splitVector command failed {errmsg:"ns not found", ok: 0.0}
[20:16:25] <BrianAnthony> looking at collections and sharding status the db my_collection and the collection collection_name do indeed exist
[20:17:48] <BrianAnthony> i'm about to just tell them to rebuild the environment since it's new anyway. something is fundamentally wrong here. i won't accept hitting an extreme edge case on the first 2 records being inserted
[20:21:29] <Neptu> hej how can i do a query with between? i want to get all the documents that are between 10 and 15
[20:40:38] <timeturner> I basically already have indexes in mongodb on what I would like to perform full text search on
[20:45:21] <_m> You could roll your own search class if you wanted to use a purely mongo solution. We've gotten better mileage with ES than internal searching classes, so YMMV
[20:47:29] <timeturner> yeah, haha. I was just worried that I would have to use a Java client as well but just found out that there is a node client
[20:49:05] <timeturner> also, I've seen the term "mongodb river" thrown around a lot with elasticsearch
[20:49:35] <timeturner> yeah I was looking at solr too, but too much config for what I need right now
[20:50:25] <_m> Pretty sure it refers to that (or something like it). Which would be an interface plugin. We don't use it here, so that's about all I can offer,
[20:51:44] <timeturner> so just to ensure I'm thinking about this correctly, elasticsearch will basically hold a duplicate of the indexes that I want to search on from mongodb correct?
[20:52:12] <timeturner> and then it uses it's algorithms to figure out which entries from the index to return based on the query I give it from the client
[20:55:32] <_m> Obviously you would need to ensure/implement that functionality; not sure how other clients handle it.
[21:10:42] <jergason> How can I have a compound unique constraint in mongodb? Does that question even make sense?
[21:42:42] <Neptu> hej have a problem, im using the primary host of my replica set and if i give the driver " replica_ip, master_ip" does not do the reading, is this normal?
[21:47:28] <pejoculant> Hi if I lose all of my config servers in a sharded set up and chunks have migrated since my last backup, am I screwed?