Log file Viewer
#mongodb logs for Friday the 31st of August, 2012
[00:49:27] <fjay> random question.. if i have a secondary mongodb cluster .. that I do a mongorestore on every day from a backup... will deletes 'casscade' when i run the resore?
[04:16:38] <brianpWins> After I get mongo installed on my server what steps should i take for securing it ?
[05:10:10] <alexluya> Hello,can anyone give a suggestion to prepare mongodb data for apache mahout(data mining tool)?
[05:23:21] <alexluya> I think,this is a real and big requirement,someone should have some experience
[07:09:35] <Gargoyle> This seems to be happening a lot since upgrading to 2.0.7 from the 10gen repo. Running ubuntu 12.04. http://pastie.org/private/i0cl5diq7wxdjwckuvgisw
[07:14:49] <Gargoyle> Looking at the backtrace, suggests it's happening when mongo tries to connect.
[07:15:35] <Gargoyle> seems to have started the other day when I switched from 2.0.4 from the std 12.04LTS, to 2.0.7 from 10gen
[07:17:32] <Gargoyle> Although I am sure I could make use of the aggregation framework if I was forced into it!
[07:19:01] <Gargoyle> so if 2.2 causes problems, then maybe I would be forced into going back to 2.0.4
[07:19:29] <ron> funny, we're also less than a week before going live, and that's why we won't upgrade to 2.2 now.
[07:19:59] <ron> almost gave our system integrator a heart attack when I told him we need to upgrade all our servers to 2.2. :p
[07:21:19] <ron> we have too many servers, too many configurations and too much other work to do to take the risk. we'll do it in a few weeks probably.
[07:21:31] <NodeX> kali : the only thing I have noticed so far is resident memory has dropped and virtual increased
[07:23:27] <ron> our would end up being write intensive with more users, but since we're not even live yet, there should be a very low initial load at first.
[07:26:14] <ron> that's right, kali, but in a week from now we're going out of stealth mode, so I should be able to link you guys to information about what we do then.
[07:27:48] <kali> we got outed two days before the end of the embargo by tech crunch. most interesting two days of my life so far
[07:42:59] <Gargoyle> I also now have "tail -f" running on the apache error.log on both web front ends.
[08:03:17] <Gargoyle> Can I stop mongo from logging connections? This log file is going to be HUGE by the end of the day!
[08:23:05] <Gargoyle> I set it in the config, and then restarted mongo. but it was a secondary node.
[08:23:32] <Gargoyle> then all of a sudden, my apache error log filled up with connection exceptions, and I had to restart apache.
[08:24:05] <Gargoyle> Should I be checkking for the exception in code, and manually triggering a reconnect - or is this something that the driver is not handling properly?
[08:33:54] <Gargoyle> Actually, that could be bad code - looks like a git merge (other developer) has removed the replica set config! :(
[08:36:53] <Gargoyle> So assuming my connection string is "mongodb://ironton:27017,kenton:27017,kirtland:27017"
[08:37:46] <Gargoyle> If I take one of those servers offline (a secondary for example), should the connection automatically switch to another one?
[08:45:02] <Gargoyle> I'll just leave it on for an hour - "Just to take the chill off" as mum would say! :)
[09:06:59] <NodeX> I just had a massive load of "Broken pipe" errors when trying to do a rather large update
[09:07:59] <Derick> Gargoyle: I'd rather have you test the new code if possible. In case you do send in a CS ticket, please try to create a real backtrace - one sec, will contact you in PM in a bit
[09:34:49] <NodeX> Has something changed in 2.2 with regards to safe writes, I have a query that writes the the DB then a function that reads it afterwards (for a ratings thing), if I read after I wrote with safe write on i get nothing returned, if I run the query again I get the expected results
[09:42:29] <NodeX> it seems that my upsert was not previously "safe" and the update to 2.2 made my query even faster so there was a race condition
[09:49:45] <Gargoyle> Page generated in 3.4020 seconds, from host hilliard, using 6,966.70KB of RAM (env: beta)
[09:50:43] <Gargoyle> most likely bits of code returning 100's of rows and only drawing 3 on the screen!
[09:51:21] <Gargoyle> That's a time from PHP code. Not even sure I want to see the browser time at the moment!
[09:52:30] <Gargoyle> PHP Fatal error: Uncaught exception 'MongoConnectionException' with message 'Cannot assign requested address'
[10:04:22] <Gargoyle> Removing virtually all calls to save($doc) in favour of update({..}, {$set:{}) calls.
[10:07:45] <Rhaven> Hi all, someone can help me with "stale config detected for ns" message that appear often in log ?
[10:42:06] <Rhaven> Hi all, someone can help me with "stale config detected for ns" message that appear often in log ? I'm using version 2.0.6
[12:45:58] <gigo1980> hi all i have a cluster with some nodes…. i have the error that my mongos process client connection block if i enter "show dbs"
[13:41:06] <Rhaven> Hi all, someone can help me with "stale config detected for ns" message that appear often in log ?
[14:18:21] <octalmind> I've got a performance issue that I would like to discuss, maybe somebody could help me
[14:21:13] <ppetermann> octalmind: in most chats people see metaquestions as rude opposed to asking what you want to know
[14:22:41] <octalmind> ok :) so - I've got a collection of 2.6 millions points with a location field containing longitude and latitude, a stopTime and a number field. I need to find the neigbors for all this points with a very limit hardware (aka my macbook pro)
[14:24:03] <octalmind> unfortunately it takes 100ms to find the neighbors for one point, and It will took me year to perform the task. I have check my indexes, but is there any way to tweak my database to speed up geo spatial query ?
[14:26:35] <ppetermann> octalmind: are you sure its mongodb thats the problem? i could imagine that you run out of ram / are swapping there
[14:30:46] <NodeX> if you want a faster query in geo spatial I suggest you use a bounding box and not a sphere as it's faster
[14:31:30] <octalmind> @NodeX Because I do not have a production yet, neither I have server yet. I am a bit constrained at my work :P
[14:32:02] <octalmind> @NodeX and I am aware that this isn't the best hardware to perform such a task
[14:33:47] <octalmind> @NodeX ok thx :) does 100ms seems to be a huge time for a geospatial query ? I am afraid to have missed something in my setup.
[14:45:19] <Gargoyle> With each level being one to many (so a category has many items, and items have many reviews)
[14:47:22] <Gargoyle> Is it possible to fetch all the reviews for a categoy (which I do save in the review), but - I need the ones for a specific item to be first, and page the results!
[14:48:55] <Gargoyle> We are talking 1000's of reviews per item, so I don't think loading them all and sorting is an option
[14:50:13] <NodeX> octalmind : the only way you're going to speed it up is to use a box not a sphere
[14:50:48] <NodeX> Gargoyle : can you explain it a bit better, I dont understand what you're trying to do
[14:52:14] <Gargoyle> NodeX: Items belong to a category. and items have many reviews. When viewing an item, I want to show all reviews for that category, but have the ones for the item you are looking for, first.
[14:59:49] <Gargoyle> I think that's going to be too slow. Thanks though, just wanted to make sure there wasn't something I was missing.
[15:00:43] <Rhaven> Hi all, someone can help me with "stale config detected for ns" message that appear often in log ?
[15:16:24] <Almindor> (cent OS 6 using 10gen repos just did yum update mongo-10gen mongo-10gen-server)
[16:15:27] <NodeX> I officialy hate all versions of internet explorer, 2 hours debugging a fricken typo
[17:29:59] <Gargoyle> It's quite possible I just didn't notice it before! Especially how mashed my head is today!
[17:30:41] <Gargoyle> cursor->count() = size of collection. cursor->count(true) = size of result set
[17:33:29] <Gargoyle> When your done building your query params array, it's helpful to actually pass it into the ->find() method! :S
[18:08:23] <Martin_> Hello guys. I'm having a bit of a problem: We just launched our social game yesterday on facebook that's using mongodb on AWS. I'm heaving a bit of a headache because we are having only ~600 player playing at the same time and ours servers are processing through the roof. But, our server is sharded in two different replica sets, and the strange part is that even though our data is split, on the servers from the first shard are bei
[18:13:46] <kali> are you sure the problem lies with a sharded collection and not an unsharded one ?
[18:23:35] <Martin_> I think it's some of the collections that are having problem... it's sharded by our internal use user id
[18:26:39] <TTimo> hello .. newb here. I'm confused why the second form in this isn't working: http://pastie.org/4641574 ?
[18:26:43] <kali> you need to identify the hammering queries and understand why the shard key choice make all of them going to the same chunk
[18:27:03] <queso> When I try connecting to a remove mongodb I get: Error: couldn't connect to server ... src/mongo/shell/mongo.js:93 exception: connect failed Could I have some help diagnosing this?
[18:28:29] <kali> queso: is the server running ? can you connect locally ? can your server have a firewall ? can your client have a firewall ?
[19:53:19] <vicelow> wereHamster, yes but the problem is there are a lot of data in this document, then If I get all the data the query is too slow I would like to do the query using $gte and $lt but I dont know how
[19:56:30] <wereHamster> vicelow: if you know how to do the query in the mongo shell, then you also know how to do it in mongoose
[19:59:30] <sqwishy> Is there a more efficient way to write this query? http://pastebin.com/CPhmSV6N
[20:27:30] <nick_h> is it possible to create an index that only includes documents that match a query? E.g. create index {foo: 1, bar: 1} on documents that match {active: true}
[20:29:23] <nick_h> vicelow: how about: db.foos.find( { "nested_level_1.nested_level_2.some_property": "some value" } )
[20:29:45] <crudson> nick_h: sparse index will be your closest bet if you want to not index every document, but it will still be on the documents for which the index is specified, otherwise it makes little sense. If you're going to query on 'active', then include it in the index, and have a sparse one.
[20:30:48] <nick_h> crudson: i considered a sparse index, but that'll still index the docs that have "active" set to false, which is a waste of RAM
[20:31:57] <crudson> nick_h: is it possible to represent 'false' for those documents by just not having the attribute?
[20:36:31] <crudson> nick_h: I appreciate that, but your primary concern re. your document structure is how it is going to be used. If it's deemed correct to have false/true for that key then you won't be able to sparsely index on it, I'm afraid.
[20:40:09] <nick_h> crudson: only 83K at the moment: http://pastie.org/private/otbf6ts9boa2iqqmn8eoya
[20:46:56] <crudson> nick_h: I hear you. those are very small documents, but it's the count obviously that impacts index size. It's good you're thinking about it in advance.
[20:48:50] <nick_h> crudson: that average size doesn't seem right, actually. here's an example doc in the collection. that's a lot more than 289 bytes, no? http://pastie.org/private/o2ao25fkyf2qeokyhk3ttq
[20:49:49] <tanob> hey guys, I'm using Ruby's Mongoid and i see them recommending some options but I can't trace them to the gem, I believe the safe options are passed to Mongo and I can't find the official doc about the options, do you know where can I find?
[20:56:03] <crudson> nick_h: Object.bsonsize(someObject) will tell you for a particular object to verify it
[20:58:10] <icenine457> can anyone explain why this query is throwing "can't have . in field names" when im clearly trying to use the positional operator? db.surveyTemplates.update({'QuestionConfigurations' : {$elemMatch: { 'Title' : 'PreferencesCollegeLevelSpanishCourses'} } }, {'QuestionConfigurations.$.Options' : 'Options:0~1~2~3~4~5~6~7~8~9~10~11+'}, false, true);
[21:07:52] <crudson> icenine457: First comment is $elemMatch with a single item makes no sense. Just do {'QuestionConfigurations.Title':'PreferencesCollegeLevelSpanishCourses'}
[21:13:28] <bcardarella> nick_h: this is on linode and they lost power, when it came back the data directory is gone
[21:13:30] <_Tristan> if I have some users, and they look something like this: {name: "Tristan", pets: {cats: 2, dogs: 1}}, can I select all users with one or more cats?
[21:13:39] <crudson> icenine457: I also think you probably want to use {$set:{'QuestionConfigurations.$.Options' :'Options:0~1~2~3~4~5~6~7~8~9~10~11+'}}
[21:14:31] <crudson> icenine457: actually it will probably try to replace the matched documents with an attribute that contains the literal '$'