Log file Viewer
#mongodb logs for Thursday the 7th of June, 2012
[00:05:34] <henrykim> is there anyway to recognize a document to load whether it is page-fault of not? I mean that loading operation is from memory or not.
[03:43:41] <deoxxa> freezey: a benchmark will tell you about the performance of one very small part of a system
[07:40:05] <wereHamster> why do we need yet another logger? This channel is already beeing logged, see the topic
[09:04:23] <negaduck> hi! how to delete an object attribute? In ruby i've got pr = $db['accounts'].find_one({:uid => uid}). I'd like to delete, for example, pr['email'].
[09:15:39] <SLNP> Are there any outstanding bugs to do with the PHP driver and connecting to replica sets?
[09:18:00] <bjori> ranman: there shouldn't be any problem - there are couple of tickets about when reading from secondaries.. but we haven't been able to reproduce it
[09:19:03] <SLNP> we are unable to connect to the set we get: PHP Fatal error: Uncaught exception 'MongoCursorException' with message 'couldn't determine master' in /root/mongotest.php:7
[09:21:47] <SLNP> test results: =====================================================================
[09:21:51] <ranman> SLNP: I have that same test failing, let me see if I can figure out what is going on
[09:22:17] <SLNP> Test for bug PHP-320: GridFS transaction issues with storeBytes(). [tests/bug00320-2.phpt]
[09:22:27] <SLNP> Test for bug PHP-320: GridFS transaction issues with storeFile(). [tests/bug00320.phpt]
[09:22:28] <SLNP> Test for bug PHP-343: Segfault when adding a file to GridFS (storeFile) [tests/bug00343-2.phpt]
[09:22:40] <SLNP> Connection strings: unsuccesfull authentication [tests/connection-with-auth-failure.phpt]
[09:22:41] <SLNP> Connection strings: succesfull authentication [tests/connection-with-auth-successfull.phpt]
[09:22:55] <SLNP> $d = new Mongo("mongodb://10.1.1.1:27017,10.132.111.43:27017", array('replicaset' => 'test1'));
[09:35:57] <bjori> SLNP: I'm wondering though, you said you can connect if you don't connect as a replicaset? what do you mean by that?
[09:39:01] <SLNP> We ran monosniff, and tested. When the replicaSet option is defined on the client we cannot see anything from the sniff output coming from the client
[09:40:48] <bjori> SLNP: please file a bug report, and include all the data you can (your script, mongod configs, rs.status()...)
[09:42:18] <SLNP> again apologies for the spam, I didn't realise my client was going to create a new message for every line. :(
[10:07:56] <SLNP> The replica set is set up using host names not IP addresses, so when we use the hostnames in the connection string it works. However this is not the case with the Python driver, with the Python driver you can pass an IP address and it's fine.
[10:43:25] <bjori> SLNP: ahh! that makes sense. the php driver uses the information retrieved from the replicaset, and if you added the servers with hostnames your machine needs to be able tor esolve it
[12:00:59] <abrkn> ec2 image http://www.mongodb.org/display/DOCS/AWS+Marketplace#AWSMarketplace-MongoDBAMI
[12:01:19] <abrkn> mongolab crashed for the 745389748592349th time today so i'll just have to set something up myself
[12:48:58] <kanzie> like collection users, I want to add field active = 1 to all existing entries in the collection
[12:53:37] <kanzie> hmm, if I have a collection in a collection then… didn't quite work. I have users and in users I have a collection called privs that has entries. I want to do something like db.users.privs.update({}, {$set:{'active':1}}, false, true)
[13:02:16] <kanzie> it seems the problem is that Im truing to create a element in a nested collection
[13:09:09] <algernon> or rather a users collection, where each document also has a pickups document?
[13:12:28] <algernon> 14:52 <algernon> kanzie: db.users.update({}, {$set: { "privs.active": 1}}, false, true)
[13:13:17] <algernon> db.users.privs does not exist, privs is an embedded property of documents inside users. so you need to db.users.update().
[13:13:17] <kanzie> NodeX algernon: that gives me can't append to array using string field name [active]
[13:22:24] <kanzie> just want to not have to delete and build the collection from scratch or manually have to update each document
[13:23:25] <kanzie> NodeX: yes, privs can have several items, each has a unique date and active-flag
[13:25:29] <kanzie> I mean if it gets to tricky Id rather just write a command to clear privs on each item in users and I can populate it again
[13:26:07] <NodeX> you can probably hack somehting using $ and dot notation but it's prolly quicker to do it this way
[13:27:55] <kanzie> Thu Jun 7 15:24:26 uncaught exception: can't have . in field names [privs.0.active]
[13:40:02] <NodeX> anyone know if 2 $inc's can be called on the same upsert ... (different fields of course)
[13:49:23] <edvorkin> Hi. I am new to MongoDB, just get installed one on EC2 from amazon Marketplace. what is the correct way to secure it? I see only port 22 open
[13:51:06] <ifesdjeen> edvorkin: limit interfaces you listen to, add authentication, disallow foreign connections
[14:38:02] <edvorkin> security question - in sharded configuration, should I add authentication only to mongos or to all members of the cluster?
[15:42:51] <railsraider> i see a lot of killcursors: found 1 of 1 and mongo loosing connections, any idea why?
[15:55:07] <m4nti5> I've been playing arround with mongodb a few weeks now using the mongo-C driver, BUT I was using an old version (v0.4), I downloaded from git the 0.6 version and saw that there is a new parameter in several functions, I've been trying to find out how that works, but I think I don't get it yet, can you recommend me some documentation that describes the behavior of that param?
[17:57:02] <crudson> It's more annoying to repost that channel's contents here; I already have all that in another tab.
[17:57:51] <carsten> logging all traffic without explict permission or option to opt out logging is rude
[18:25:50] <mbarr> Hello all.. trying to determine the impact of a large /dev/shm on Mongo's use of memory - RHEL 6.2 x86_64 linux...
[18:26:50] <mbarr> I don't believe it's going to use the memory, but i'm just checking to make sure it's not able to shrink it automatically, or use it in another fashion.
[18:29:03] <tomlikestorock> how do I query on a key's value that is in a list of embedded documents?
[18:50:54] <fkefer> Hi all; I tried installing mongod-gen10 server in fedora17, but selinux insists on not allowing it
[19:02:33] <m4nti5> tomlikestorock: please be more especific with what you want, perhaps use paste bin to paste an (idented) example of the document and say
[19:14:46] <m4nti5> is it possible to set a mongo_write_concern in a mongo_run_command function in the mongo C driver?
[19:20:50] <ranman> m4nti5: http://api.mongodb.org/c/current/write_concern.html?highlight=write%20concern
[19:24:53] <m4nti5> ranman: thanks for the link, I've read it already, and didn't say anything about a way to use it for the mongo_run_command function, I also read http://www.mongodb.org/display/DOCS/getLastError+Command, my question is actually can I do a getlasterror all along with the mongo_run_command
[19:28:29] <pneff> Hey I need some help with a query. Im running an or query that contains two different fields and is sorted by a date field. When I run each part of the or segment individually or without the sort everything runs fine but when i try to run the entire thing as one query it stalls out. I'm assuming this has something to do with the indexing. Each field that is searched or sorted has an index but not a composite index. Here's my query : db.cdr_record
[19:28:29] <pneff> s.find({$or:[{called_num:"7033483543"}, {trans1_num:"7033483543"}]}).sort({start_time:-1}) Any reason why this would just stall out?
[19:53:50] <harrymoore> pneff: not sure this helps but mongo will use at most 1 index for a query. so you may need a composite
[19:55:01] <mattbillenstein> anyone know of a good way to dump/restore a database over a pipe — like over ssh?
[19:55:41] <pneff> but I'm only selecting on one since its an or query. http://www.mongodb.org/display/DOCS/Indexing+Advice+and+FAQ
[19:59:14] <rudolfrck> anyone here using mongo_mapper with rails 3.2? I'm looking for some help on a small issue...
[20:06:32] <pneff> harrymoore: but the query works just fine if its only a single field with the sort which is kind of annoying
[20:09:16] <harrymoore> pneff: try a composite index which also includes the field you are sorting on. try .hint() to force the query to use that index
[20:25:38] <mateodelnorte> Anyone have pointers for debugging why a node app would "crash" sans error message or stack trace while interacting w/ mongod? The only clue I have is seeing "SocketException handling request, closing client connection: 9001 socket exception [2] server [127.0.0.1:51434]" in my mongod logs. Using node-mongodb-native.
[20:49:37] <dstorrs> you have two records with the same ObjectID value and you want to change one of them?
[20:50:56] <dstorrs> does anyone have experience using Mongo as a backing store / intermediation point for a distributed messaging service (a la RabbitMQ, Celery, etc) ?
[20:51:36] <m4nti5> no, I'm using the mongo C driver, and I create a bson in the client side with an ID (a regular bson_oid_t), the thing is, there is a chance that the ID I'm generating it's already saved in the server (because all ID's I'm generating are client-side IDs)
[20:53:12] <m4nti5> so that will generate a duplicate entry error, the thing is, the bson migth be too big so copying it or regenerating can be very heavy, so, I was wondering if somebody knows a way to re-generate the bson_oid without creating the bson again
[20:54:09] <jmorris> would anyone know why the output of Model.find(...) in mongoose would not agree with what i see from the command line mongo : db.model.find() ?
[21:09:16] <wb_> when storing images on s3 or gs and managing meta-data in mongo, do people typically create a separate collection of image meta-data? or embed the meta-data inside the documents they are associated with
[21:10:15] <wb_> it seems like it would be easier to just have a separate collection for image meta data, rather than constantly reaching inside documents.
[21:11:20] <dstorrs> wb_: if the metadata is finite and not constantly growing, I would probably embed it. but either way works
[21:12:07] <dstorrs> in general NoRel DBs (a term I prefer over 'NoSQL' -- more accurate) reward you for having a rich data model and putting it all in one document
[21:13:49] <wb_> thanks dstorrs. I think I will try just sticking it in one document, and hope for the best
[21:34:41] <themoebius> I'm getting these messages on my mongos server. ChunkManager: time to load chunks for pb3.hourly_stats: 2990ms sequenceNumber: 1425 version: 38915|169
[21:35:16] <themoebius> I've manually moved a few chunks under high load i wonder if it has anything to do with it
[22:55:35] <richardraseley> Question: Can anyone point me towards resources that talk about the multi-datacenter capabilities of MongoDB?
[22:56:40] <richardraseley> I am primarily interested in finding out about how we can read from and write to the same data from multiple locations - depending on optimistic replication to reconcile the differences over time.
[22:58:36] <richardraseley> Also, does MongoDB provide the capability to scale the active capability of a particular shard. It seems (based on my limited understanding) that only one server in a replica set will ever be able to handle both reads and writes, and therefore you are limited in shared performance for writes to that of a single server. Is the solution to this simply creating more shards across which you will spread the entire data set?
[23:05:06] <m4nti5> when you're using the write_concern attr, it internally calls to mongo_check_last_error, and there the response bson is not destroyed
[23:05:49] <dstorrs> richardraseley: http://docs.mongodb.org/manual/administration/replication-architectures/ Look for "Geographically Distributed Sets"
[23:11:17] <m4nti5> dstorrs: hummm I'm kind of lost here, it looks like that link is for server reports
[23:11:19] <richardraseley> dstorrs: In that scenario, a second site can always only provide reads / DR.
[23:11:44] <richardraseley> dstorrs: There can never be two active (providing writes and reads) sites such as you could have with Cassandra, correct?
[23:25:12] <richardraseley> dstorrs: How so? The documentation seems clear that you can only have one master in a replica set... or am I not understanding that?
[23:26:04] <richardraseley> So, even if I have 3 shards, two of which have masters in site A and one in site B, I can still only fully access data (on whatever shard it lived) from one site at a time.
[23:27:44] <dstorrs> you said "there can never be two active" -- I understood that to mean "servers". which is perfectly fine if they are masters of separate shards.
[23:28:49] <dstorrs> as to doing multi-master replication throughout a replica set -- I believe Mongo may have supported something like that at one point, but they migrated away.
[23:28:59] <richardraseley> Sorry, that was not my intent. To put it differently, one can never have read / write access to the same chunk from two different physical sites and expect those operations to be eventually consistent.
[23:29:07] <dstorrs> maybe it was just regular master / slave I didn't care enough to research it since it wasn't relevant.
[23:30:16] <richardraseley> Heh, that is why I am in this channel. =] I appreciate your help though.
[23:30:23] <dstorrs> I'm able to answer most of the questions in here because I've actually RTFM'd a lot, but I haven't had the chance to do replication yet, so it's just book learning