Log file Viewer
#mongodb logs for Saturday the 16th of June, 2012
[00:13:07] <mkmkmk> i have another box up that had mongos running before the 3rd config died & was brought back… it's operating fine
[00:24:25] <mkmkmk> if mongos dies on another box and wont come back, i'm going to be dead in the water
[01:54:13] <Kryten001> Hi, here http://www.mongodb.org/display/DOCS/Object+IDs they say that an _id must be unique for a given db cluster, does that mean that two objects in the same db but different collection can not have the same _id ?
[04:34:07] <dstorrs> I have a collection that stores statistics. When I add new stats, I would like them to include incremental data (current stat - previous value). Is a mapReduce the best option here, do you think?
[04:34:54] <dstorrs> Doing it one by one is certainly out, but is there an alternative to M/R that I haven't considered?
[05:48:01] <oktapodi> how should i create the relation ship? should i embed both in "Band" or should i create a different relationship for each?
[06:12:00] <dstorrs> you've got two options -- put each entity in a different collection and do the "joins" in app code
[06:17:20] <heoa> What was the MongoDB -community-blog where blog-posts were fetced from the git repo?
[06:50:54] <dstorrs> oktapodi: you're welcome. (pardon long ping, I'm still working and a bit distracted)
[13:34:11] <mitsuhiko> does anyone know if i can make a mongos just die on error instead of trying to recover itself?
[13:34:33] <mitsuhiko> if a mongos comes up before a replica set then the whole thing just zombies out and never recovers by itself
[14:09:02] <mitsuhiko> it's neither reliable nor easy to use nor well designed. but compared to many other nonsql solutions it has some merits
[14:09:03] <Defusal> which brings me to the point where i need to start deciding whether continuing to invest in mongo for projects is worth it in the long run
[14:11:13] <Defusal> i want the advantages of a non-linear schema, although embedded documents arnt always great, they work well in certain cases
[14:11:39] <Defusal> and having to constantly update the database schema like i do with projects using SQL databases, is just a pain in the ass
[14:12:30] <Defusal> it should be up to the developer to take care, not the database to force unnecessary action
[14:15:16] <Defusal> while there are some downsides to the query interface, for the most part, the schemaless nature does streamline development, especially for prototypes
[14:16:19] <Defusal> i cringe when i have to make a single change to a project using a SQL database just because of the effort it takes to update the schema
[14:17:01] <mitsuhiko> Defusal: the problem with mongodb is not mongodb itself as much as the fact that everybody thinks he needs to use it
[14:18:10] <mitsuhiko> most of the problems that people complain about are happening with zero or one config server
[14:18:17] <Defusal> although a platform i designed when i was contracted out to do so has grown to such a point since i left
[14:19:14] <Defusal> well, is there an article somewhere pointing out the right setup if you want to avoid data loss?
[14:20:20] <mitsuhiko> even the official docs pretty much tell going from non-sharding to sharding is painful
[14:20:43] <Defusal> very small setups (community projects) work fine, and would not need to be scaled up much
[14:21:01] <Defusal> some of my other projects like my current one will definitely be scaled up in the future
[14:22:12] <Defusal> if a single server works fine as is, is the threat of data loss really large enough for it to be worth a sharded setup
[14:24:23] <Defusal> a replica server would obviously be better protection than backups every few hours, sure, but how often is there gonna really be a data loss issue that will require recovery?
[14:26:08] <mitsuhiko> Defusal: mongodb is just fine for what it is. but you have to take it with all it's warts and bugs :)
[14:26:20] <mitsuhiko> i think the argument for mongodb is that 10gen is very, very active and cares abuot improving their product
[14:26:36] <Defusal> mitsuhiko, well then again, please point me to an article that provides details on this
[14:27:56] <Defusal> but as i said, you are the first one to ever tell me a replica set is *required*
[14:28:18] <Defusal> i have never seen anyone, on irc, or an online article, say that is a *requirement*
[14:29:38] <Defusal> but now that mitsuhiko says this, i would like to know what the real world chance of data loss without a replica server is
[14:32:04] <kali> Defusal: i mean, you're alwaus at the mercy of a catastrophic hardware failure, and these are not that rare
[14:32:25] <kali> Defusal: so if you are willing to risk having to start up from the last backup, yeah, why not
[14:35:54] <mitsuhiko> Defusal: because you need to introduce a mongo controller, mongo routers and the balancer will spend a long time moving your data over :)
[14:37:31] <Defusal> well, i guess it may very well be easier to create more seperation when scaling up
[14:38:53] <mitsuhiko> anyone any ideas how to make a mongos die instead of retry itself to find the config server?
[14:39:10] <mitsuhiko> if a replica set is entirely down and the mongos is up it will never recover from it :(
[18:24:13] <skot> not really related as varnish is an http (mostly) accelerator and mongodb is a document database.
[18:24:51] <AlecTaylor> But does Varnish use a db? - I'm sure it would... can it be modified to use Mongo?
[18:25:29] <skot> It uses in-memory/persistent data structures but also use memory mapped files for those
[18:30:30] <kali> nope varnish accepts modules for request processing, but it has only two builtin storages: one mmap base, the other malloc based
[18:33:00] <skot> cool, thanks for the info, do you happen to know how many use the malloc based version; I thought the mmap one was the default.
[18:35:07] <skot> with what I've seen with shared server running mongodb, and the mmap os behavior, I can imagine it would not go well using that model with varnish
[18:36:07] <kali> well, it make sense to have something like a haproxy or an nginx, as they are not memory greedy
[18:54:28] <jpnance> let's say i have a collection of documents such as: { "field1": 10, "field2": 5 }
[18:54:45] <jpnance> what's the typical way to query that collection for all documents which have field1 > field2?
[18:55:18] <wereHamster> which is slow, because it can't use an index. If you want to use an index, make a third field, field1minusfield2, and index that. and then run a query on that
[20:43:45] <kaikaikai> i'm running into this problem using php drivers/library: http://jsfiddle.net/haSYw/2/
[20:43:59] <kaikaikai> the error is "MongoCollection::update() expects at most 3 parameters, 4 given"
[20:44:25] <kaikaikai> but to me, the update command i'm running seems well formed according to docs http://www.mongodb.org/display/DOCS/Updating
[20:46:04] <kaikaikai> i mean, docs specify four arguments in the introduction, so how could it expect at most three?
[21:14:52] <kaikaikai> ahhh, thanks skot and toothr, thought the mongo docs were a general outline for all drivers
[21:16:41] <toothr> iirc there are actually links to the related driver docs for the popular langauges on each doc page