Log file Viewer
#mongodb logs for Wednesday the 15th of April, 2015
[06:26:54] <al-damiri> I'm new to MongoDB. I type `mongo` in my terminal and it says connecting to test.
[06:27:20] <al-damiri> When I do `show databases`, it shows local and one another database but not test.
[06:32:21] <Boomtime> you have connected to the 'test' database, but it is ephemeral until you do something with it
[06:32:44] <Boomtime> once you insert a document, or create an index, or apply some other persistent action, it will show up
[06:40:13] <al-damiri> Another question if I only type `mongodump` will it backup the all the databases (with users/roles/passwords etc)?
[06:43:38] <joannac> al-damiri: yes, as documented here http://docs.mongodb.org/manual/reference/program/mongodump/#cmdoption--dumpDbUsersAndRoles
[07:21:01] <al-damiri> joannac: Oh, I see. Thanks. Let me see then how can we create a user. Thanks :)
[08:41:02] <grepwood> I'm doing `mongo --host localhost --port 27017 --eval "rs.conf()"` and I don't get anything useful
[08:41:25] <grepwood> and I can't find anything helpful in the documentation apart from writing javascript files and then feeding them into mongo
[10:25:52] <fontanon> Hi everybody, is it possible to configure a cluster of 2? Imagine I have two hosts (server1 and server2), this is what I get when I try to launch mongos: "BadValue need either 1 or 3 configdbs" $ mongos --configdb server1:27019,server2:27019
[10:33:50] <fontanon> so, in order to have two shards in separated servers, I would need an additional server to use as the third config server
[10:34:18] <Derick> well, in production, each shard should have two datanodes *and* either another data node or an arbiter
[10:39:54] <fontanon> Derick, but I'm not trying to replicate data in different machines but setting a 2-tier schema with tag aware sharding: http://blog.mongodb.org/post/85721044164/tiered-storage-models-in-mongodb-optimizing
[10:40:30] <Derick> http://docs.mongodb.org/manual/core/replication-introduction/ has a good introduction about replica sets though
[10:41:00] <fontanon> Derick, I expected to run that on 2 machines, one with ssd for recent records, another with massive hdd for older records.
[10:42:23] <Derick> you really should have replication though for failover, because if a node goes down - so does your app
[10:43:31] <fontanon> Yes, on DO, is a cloud provider, forget what I said about my own data centre (I didn't understand the question).
[10:44:04] <Derick> in production, you really should use a replicaset (one for each shard), unless your data isn't important
[10:45:06] <fontanon> Understood. So 4 machines would be a good schema? 2 for the shards and another 2 for the replicaset ?
[10:45:55] <Derick> so two shards, means you have two replicasets. Each replicaset should be 3 nodes (2 data + 1 arbiter, or 3 data)
[10:53:19] <Derick> but you can probably get away by putting them on the same physical machine as an arbiter
[10:54:36] <fontanon> then I didn't understand your last comment: "<Derick> but you can probably get away by putting them on the same physical machine as an arbiter"
[10:55:50] <Derick> that's a "mongos" node, which you typically put on the machine that runs your application code
[10:58:41] <Derick> you can either pick: 3 data nodes (mongod), or, 2 data nodes and one arbiter. An arbiter is a lightweight process, that does not store data - but it does help out in case a data carrying node goes down
[10:58:57] <Derick> they're basically a judge to say which of the two data nodes is the "primary" node
[11:00:31] <fontanon> doesn't mind I reached here: http://docs.mongodb.org/manual/tutorial/add-replica-set-arbiter/
[11:00:56] <Derick> you start it up as a normal mongod node, but add it to a replicaset set with addArb isntead of add
[11:11:58] <KekSi> also start it with --smallfiles and --nojournal and the one you should connect to to use rs.addArb() is the PRIMARY of the rs
[12:06:50] <hernil> Hi! I'm quite new to nosql databases and I'm trying to let go of my relational reasoning when modeling my application. I'm building a system with some similarities to Stack Overflow and I'm thinking about tag handling. Should I just keep them in the "question" document or should I normalize it? My main problem with not normalizing is performance concerns regarding searching through all tags (suggestion of tags). Will it be considerable or am I overthinking it
[12:18:28] <StephenLynx> but then theres indexes and I am not sure how that could be optimized, searching for documents that contains an element on an array.
[12:19:47] <StephenLynx> I don't think even SO looks for all posted questions in search of suggestions.
[12:24:45] <hernil> StephenLynx: How would that look? Is this something I would do myself or is this abstracted away by using mongoose as I am now? :)
[12:25:04] <deathanchor> you can also keep a tag collection (indexed by tag) and just store the object id of the question from a questions collection
[12:25:31] <deathanchor> but that's two lookups and you have to maintain both collections to keep them in sync
[12:42:39] <deathanchor> what op should I use if want to do a find() where 'string' is matched at the beginning of value?
[12:44:48] <StephenLynx> don't know how they handle indexes neither, but the documentation should give you that information.
[13:41:43] <deathanchor> anyone know how to write the /^pattern/ equivelent for use with Mongo driver for PHP?
[13:44:09] <Croves> I don't know exactly what's the error, but when I remove 'birthday', it runs fine
[14:12:19] <deathanchor> my php mongo query not returning results: help? https://gist.github.com/deathanchor/4b0e1bc5deb4d77e5ce1
[14:23:00] <GothAlice> deathanchor: I don't PHP, but what exactly is $dayDate; when pasting code it's a good idea to include enough to reproduce. ;)
[14:31:16] <GothAlice> deathanchor: Have you tried the longer-form $regex? (I'm not actually sure if the longer form is compatible with $in usage, though.)
[14:31:36] <GothAlice> deathanchor: A la the first example on http://docs.mongodb.org/manual/reference/operator/query/regex/ (vs. the second example)
[15:02:09] <fontanon> Hi everybody, I would like to change the ports in my shards configuration, i.e. shards: { "_id" : "shard0001", "host" : "server1:27017" } TO shards: { "_id" : "shard0001", "host" : "server1:27018" }. May I have to remove the shards and add them again?
[15:05:29] <GothAlice> Since the set name isn't changing, and in theory that replica is already up-to-date, the catch-up process when it re-joins should be fast.
[15:05:53] <fontanon> GothAlice, thanks I'm removing the shards with db.runCommand({removeShard:"shard0001"}), is it ok?
[15:06:18] <GothAlice> Yup, that's the recommended approach as documented here: http://docs.mongodb.org/manual/tutorial/remove-shards-from-cluster/
[15:06:19] <fontanon> the first shard I removed with that command displays a "drainning: true" status
[15:06:29] <GothAlice> Just remember to do one node at a time, if you're wanting to change multiple.
[15:07:02] <fontanon> yep, mongo complains if I try to remove another shard if the previous has not finished
[15:07:55] <GothAlice> What you're waiting for is for all of that data to get shuffled to other nodes to ensure your data is always accessible.
[15:08:25] <GothAlice> The amount of time it takes will depend on the amount of data that needs to be moved.
[15:09:49] <GothAlice> Anything else showing up in the logs? And what's the output of db.currentOp()?
[15:10:51] <fontanon> but stills drainning --> shards: { "_id" : "shard0000", "host" : "server1:27017", "draining" : true } ...
[15:11:58] <fontanon> GothAlice, I understand this un-clustering procedure is prod-env friendly, but just because I'm playing in a sandbox ... isn't there an easy way to destroy the cluster?
[15:13:08] <GothAlice> The contents of your mongod's data directory, typically /var/lib/mongodb/*, /var/lib/mongod/*, or /data/* (if not otherwise configured, I believe this is the default)
[15:26:39] <deathanchor> I have 4 member set, m4 is zero votes and priority. would a force re.reconfig() on m4 to make it a solo member of the set break it out of the other 3 cleanly (and not affect the 3)?
[15:26:40] <fontanon> GothAlice, cheeser it was a matter of removing the config servers metadata directory for removing the cluster, no need to remove the data in the nodes, thanks for your help
[15:29:07] <GothAlice> deathanchor: The process would be to remove that member from the set using the management commands. It will still think it's part of the set, though, and will need to be restarted after altering the configuration to remove the replSet configuration option. Only _then_ will it think it's standalone.
[15:29:49] <GothAlice> If it's a non-voting, non-electable member of the set, the set should be otherwise uninterrupted during removal. (You _might_ trigger an election, which will cascade down to the client drivers needing to reconnect, but the same primary should win.)
[15:45:09] <deathanchor> yeah, I'm trying to figure out a way to take prod data to a testing env, and bring it back again to resync from scratch when needed.
[16:19:35] <zunkree> I have a question. When I dump database with mongodump they used less space than in mongodb stats. Is this ok?
[16:23:00] <GothAlice> It only gets dumped if you explicitly request "point-in-time" backup from mongodump.
[16:25:33] <GothAlice> zunkree: MongoDB on-disk data files are "sparse", i.e. they develop holes, and are over-allocated vs. the amount of actual data. (They also can become fragmented enough to warrant compaction.) Additionally, each record has some "head room" added to it to allow it to grow without needing to be moved—this helps reduce fragmentation. When you use mongodump it pulls out each record and packs them as tightly as possible in the resulting .bson
[16:26:31] <deathanchor> zunkree: I usual do a test import and check that count is the same as when I did the dump
[16:27:28] <deathanchor> I have one DB where I rotate collections which is 400x larger than the actual data size
[16:29:09] <GothAlice> deathanchor: http://www.devsmash.com/blog/mongodb-ad-hoc-analytics-aggregation-framework < this article describes several different ways of storing the same data and the impact on query performance and data storage size. Simple changes can have substantial effects on disk utilization.
[16:30:00] <GothAlice> (The naive approach, document per event in this example, stores 133mb of data and 94mb of indexes. The hybrid approach stores 39mb of data and 23mb of indexes… and it's the same data!)
[16:30:51] <deathanchor> GothAlice: my issue was we got lazy and the rotation stopped working. So 3 months of data got stored instead of 2 days worth
[16:38:30] <GothAlice> FAQ updated to include mongodump: https://gist.github.com/amcgregor/4fb7052ce3166e2612ab#manual-dump
[16:39:54] <GothAlice> We may or may not have answers, but before any can be provided the question must be posed, Tinuviel.
[16:41:11] <Tinuviel> I have mongoDB with two shards (three replica members each). Database is getting constantly some reads from web app. I need from time to time some heavy writes to DB. It will take maybe even few days, script will be reading some documents from DB,calucateing something and writeing back to DB>
[16:41:46] <GothAlice> Axy: http://docs.mongodb.org/manual/core/index-text/ < If you want to perform full text search, you're going to need to create a text index. However, these searches are restricted to a single collection at a time.
[16:42:00] <Tinuviel> I was thinking about redirecting reads to to secondary member, and do writes on primary
[16:43:01] <Axy> GothAlice, can I send you an example entry over dm? maybe there is a better way of doing this
[16:43:43] <GothAlice> Tinuviel: Alas, all writes _must_ pass through a primary. Handing your reads to secondaries will help reduce the overall read load on the primaries, but your application should also be pulling data from the secondaries where possible. Your main trick will be to rate limit your updates. If you are performing those updates all in one thread, you might not need to due to the natural serial nature of an arrangement like that.
[16:44:33] <GothAlice> The only other concern for a long-running query like that is the need to handle the cursor timing out. I.e. retry from where you left off in the event of the cursor being closed.
[16:46:25] <GothAlice> Adding a small time delay (i.e. time.sleep(0.1) in Python) to the bottom of your loop would work, as one approach.
[16:46:37] <Axy> GothAlice -- when I do db.collection.findOne() this is a single entry that comes up http://pastebin.com/QxASwpVL
[16:49:00] <Tinuviel> But can I somehow force that without adjusting query? I'm asking because query will be in PHP code, and if we will need to changed that to ask secondary, it will require to do the release, which take long time.
[16:49:36] <GothAlice> Tinuviel: That's called the "read preference". You'll have to examine the driver docs for the implications to your code.
[16:50:37] <GothAlice> In *theory* you can run a middleware server which transforms the data coming out of your app before it hits mongod, but, uh, that's a bit of overkill.
[16:52:21] <GothAlice> You could, but secondaries are potentially delayed. You may have situations where you .update() something, then .findOne() the updated copy… if the findOne() asks a secondary, you may get the document prior to the update.
[16:52:57] <GothAlice> So anywhere you think you want to use a secondary read preference, consider: is it A-OK if this query gets old data?
[16:53:49] <GothAlice> http://docs.mongodb.org/manual/reference/read-preference/ goes into depth on the various caveats of each choice.
[16:57:10] <GothAlice> Axy: find() will return a cursor which has no particular limit; internally it requests records in batches. (Each new batch is referred to as a "getMore" operation.)
[16:58:07] <GothAlice> However iterating the cursor will keep that little fact from you, excluding a slightly increased delay every 20 records.
[16:58:35] <GothAlice> Aye. In Python if you cast a cursor to a list (list(cursor)) it slurps all the data in.
[16:59:04] <GothAlice> (This is generally not a good idea, though. It requires enough memory to store all data, and blocks until all the data is read.)
[16:59:37] <Axy> Ah this is an action I'll do manually -- I jsut need to find a piece of text that's in the database,since I can't do a serach for it I was just thinking, console logging thee whole database and just manually performing a ctrl-f
[17:01:46] <GothAlice> db.foo.find({$or: [{"post.photos.alt_sizes.url": /bob/}, {"post.photos.original_size.url": /bob/}]}) < for example
[17:07:30] <GothAlice> Because you're storing the original separate from the scales, you effectively need the combined set of two queries, thus the $or.
[17:26:53] <fullstack> In Mongoose Populate (show nested), how do I show all without passing the specific column name? When I call mongoose.model...find().populate().exec(function(data){console.log(data}) I get an error
[18:46:27] <MacWinner> StephenLynx, Why not use mongoose? sorry, just noticed the post... i actually just was playing aroudn with it today.
[18:47:09] <StephenLynx> because it is bad designed, has slow performance, adds a failure layer, it completely changes the workflow, all while offering zero functionalities.
[18:47:55] <MacWinner> StephenLynx, my main use for it is adding model methods... I'm open to using something else.. any suggestions?
[18:47:58] <GothAlice> MacWinner: It is the single most frequent source of support incidents I assist with, typically caused by Mongoose encouraging patterns of use MongoDB doesn't cope well with. A typical example is the naive approach of storing ObjectId instances as strings (more than doubling the storage requirement, breaking query/sort capability, etc.
[18:49:09] <StephenLynx> cheeser and I know lots of folks that think that windows is A-ok for a server.
[18:49:10] <MacWinner> StephenLynx, so just creating a plain old javascript object and having it reference mongodb directly?
[18:50:22] <GothAlice> "Badly designed": In what ways? "Slow performance": do you have benchmarks to back that claim up? "adds a failure layer": that's minor; often the benefits of a DAL outweigh the risk of an additional moving part. "Completely changes the workflow": Not completely; it's still CRUD. "Offering zero functionalities": patently false if you look at the documentation.
[18:51:20] <MacWinner> could you point me to an example of using the native driver, but extending the document with a method like if it's an Animal object, give it a method to bark()
[18:51:21] <GothAlice> Those gross generalizations would, in theory, apply to any DAL. I use MongoEngine. I'd prefer to not work without it. (I make extensive use of its features, including automatic caching of values associated with references, signal handlers, cascading deletes, …)
[18:51:32] <StephenLynx> and of course it will degrade performance because it adds and abstraction layer.
[18:51:55] <StephenLynx> and AFAIK it doesn't enable you to do anything that you couldn't already do with the driver.
[18:54:50] <fullstack> they wanted to be able to change the name of a data item and have it reflect to all the other areas of the documents without updating the entire collection
[18:55:36] <StephenLynx> you added a dependency to make your database work exactly the oposite way it should work instead of using a more apropriate database?
[18:55:46] <GothAlice> I do exactly that to store single-character fields in MongoDB, but have descriptive names at the application layer.
[18:56:43] <fullstack> StephenLynx, they didn't want to write one off REST queries for each possible combination of changes
[18:56:51] <GothAlice> There's also "active record" (i.e. the document can save modifications to itself amongst other things), query transformation (build up queries using a more "native" approach than $ keys), etc.
[18:58:05] <fullstack> So like, active record , I'd say "Hey this field change. go everywhere in the collection and find where we used the same field where the data is duplicatd and update it. Also lets pay employee Bob to do this fulltime"
[18:58:10] <StephenLynx> you have an object that holds a reference to the document and make it perform the query it is referenced to.
[18:58:15] <GothAlice> MongoEngine is capable of "dirty flag" tracking and can transform local modifications into efficient $set/$push/etc. atomic operations.
[18:59:15] <GothAlice> fullstack: A Document Object Mapper and Active Record system would let you change the name of the property in your application independent of the name of the field in the database.
[18:59:50] <fullstack> Oh I just wanted to change the value of "name" and have it reflect everywhere else
[19:00:14] <fullstack> as in, name being a proeprty in the document. not change the name of the property
[19:00:44] <GothAlice> fullstack: So, like, db.collection.update({_id: ObjectId(…)}, {$set: {name: "Bob Dole"}})
[19:00:56] <StephenLynx> in the end, it went to all collections and updated them all. you just didn't see it.
[19:01:11] <fullstack> GothAlice, yeah but now we need to change it everywhere else that I use that property.
[19:01:41] <fullstack> in other documents. Like reports, or etc. So DBRef/Mongoose was the logical step
[19:02:16] <GothAlice> fullstack: So, db.other_collection.udpate({'some_reference.id': ObjectId(…the thing updated before…)}, {$set: {'some_reference.name': "Bob Dole"}}, multi=True)
[19:02:29] <fullstack> "name" was just one of them, First name last name, age, questions, answers, sections, teams,
[19:02:53] <fullstack> coach name. If I change the coach name for a team, I'd have to go to each player and change their coach name property
[19:03:01] <GothAlice> Aye. You're basically looking for: http://docs.mongoengine.org/apireference.html#mongoengine.fields.CachedReferenceField
[19:04:23] <StephenLynx> the problem with these featureless abstraction tools is that they are still performing the tasks that you don't want to write yourself. but with much more increased complexity because they are way much more generic than you need..
[19:04:25] <fullstack> its not immediately obvious to me how that feature would help me, but thanks.
[19:04:32] <GothAlice> AFIK Mongoose doesn't yet have support for this type of automatic caching, though in theory it is implementable using its "middleware" feature (trap "save" on referenced schemas, update references manually as described above).
[19:06:30] <GothAlice> fullstack: CachedReferenceField('Coach', fields=['name']) — if this is defined in your schema that references a coach, as an example, MongoEngine will _automatically_ handle storing the name field of the referenced document, and automatically update the cached value in the reference if the target document is updated (via MongoEngine).
[19:07:22] <GothAlice> Additionally, the MongoEngine query planner allows for treating these cached values as if they were a join: Team.objects(coach__name="Bob Dole") < actually works
[19:10:50] <saml> after upgrading mongod, mongoimport is upgraded.. and it sometimes insert as long.. sometime as int
[19:11:14] <GothAlice> saml: http://docs.mongodb.org/manual/reference/mongodb-extended-json/#numberlong
[19:11:59] <GothAlice> That means all numbers are 64-bit IEEE double precision floating point, with 54 bits of integer accuracy.
[19:12:47] <GothAlice> I suspect it'd attempt to find the most compact viable representation for any given number.
[19:15:25] <saml> play.api.Application$$anon$1: Execution exception[[ClassCastException: java.lang.Double cannot be cast to java.lang.Integer]]
[19:15:32] <GothAlice> http://bsonspec.org/spec.html — int32 (type \x10), int64 (type \x12), double (type \x01) — these are the available number formats in BSON. Of these, JSON only directly supports doubles.
[19:19:06] <StephenLynx> your problem is not that java isn't optimized for web servers like other technologies
[19:23:59] <StephenLynx> I've just finished my forumhub using mongo and io. I just need to stop being lazy and get a token in my bank to be able to buy stuff with my credit card on the internet again so I can host it.
[19:24:01] <GothAlice> MacWinner: Could you run one line for me and tell me the result? typeof "foo" === typeof new String("bar")
[19:24:08] <MacWinner> GothAlice, I've become a huge fan of javascript through a reverse route.. It started when I began with angularjs develepement
[19:25:02] <GothAlice> Welp, it _wants_ me to like it, but continues to do things that make sure I never will. ;)
[19:25:54] <GothAlice> StephenLynx: Does one still need to memorize a page and a half of hierarchical typecasting rules?
[19:26:41] <GothAlice> StephenLynx: Any time you use == instead of ===, you're hitting the casting rules. How do you cast to a number?
[19:28:34] <MacWinner> StephenLynx, wraps generator functions and makes them extremely simple to do async code in a sync fashion.. no callbacks or .thens
[19:28:35] <saml> so.. i could downgrade mongoimport, rewrite mongoimport (read json and do mongo update)
[19:28:38] <GothAlice> saml: The older version was not conforming to the actual JSON specification, in that case.
[19:29:08] <saml> probably someone opened jira ticket about mongoimport and json number.. and it got fixed for him.. but broke for me
[19:29:51] <GothAlice> saml: Yup. If you want something else, in this case, I'd recommend dumping to CSV. Then you can have explicit control. JSON just _does not do integers_.
[19:29:57] <StephenLynx> never had any issues with callback based code and debugging when I had to find where something is going on.
[19:29:59] <MacWinner> no promise chains needed.. try/catch scope nicenes.. tj hollowaychuck basically started this as a replacement for express..
[19:30:45] <GothAlice> saml: (The tears of thousands of crypto developers have left a bit of a stain on JavaScript, because of how it only does floats.)
[19:30:53] <MacWinner> StephenLynx, give it a look though.. I'm a newb with node, but after playhing around with old ways of doing stuff and this ES6 way, it's really game changing for me
[19:37:17] <MacWinner> TJ Holowaychuk and Jonathan Ong are the 2 main authors.. they are moving to this from express.. i think there is something here. i have personally found it useful, but hard for me to articulate exactly why
[19:37:31] <GothAlice> StephenLynx: My web framework's core is 649 SLoC… small is good. Small is testable. Small is easy to document.
[19:39:01] <StephenLynx> and express is awful and I have no idea who these two guys are, saying they are moving to koa from express means jack to me.
[19:40:08] <GothAlice> (My own packages record dependencies with "soft" version pinning, i.e. "less than 3.0" type stuff; I can do this on the marrow dependencies because I guarantee minor and rev bumps don't break things.)
[20:10:11] <MacWinner> noticed this when i startup mongo 3.0.2.. WARNING: /sys/kernel/mm/transparent_hugepage/enabled is 'always'. if I change it to recommended 'never', what kind of impact could that have on ther apps?
[20:19:49] <Ramone> hey all... anyone know if there's a way to cycle pooled connections in the node.js driver?
[20:20:34] <Ramone> I'm having a production issue that redployment seems to fix, and I'm wondering if it could be related to stale connections or something
[20:35:27] <pagios> hi all, trying the ruby binding and getting an argument error: result = client[:'userTable'].find({"token" => token}, :fields => ["arn"]) thats like in the doc https://github.com/mongodb/mongo-ruby-driver/wiki/Tutorial
[20:47:36] <mrmccrac> anyone know if using mongo 3.0 + wiredtiger + snappy compression if there is a minimum storage size allocated for a single document?
[20:58:21] <GothAlice> mrmccrac: Around 15 bytes or so for an "empty" document, plus padding factor. (See the docs/configuration for this value; I don't seem to have it on hand.)
[20:59:36] <mrmccrac> trying to figure out why storage size estimates aren't matching up with a larger data set based upon what i saw in a smaller one
[21:00:34] <GothAlice> Always handy to have on hand when doing napkin calculations for size: http://bsonspec.org/spec.html
[21:04:21] <GothAlice> Compression, of course, makes the sizes highly variable. (Compressing the same bytes twice may not even result in the same compressed byte stream!)
[21:12:38] <mrmccrac> i had a dataset that was 1480G, imported into mongo w/ compression on one host and it ended up being 173G
[21:12:40] <GothAlice> mrmccrac: You seem to be missing a qualifying word between "lot" and "disk". ;)
[21:13:55] <GothAlice> That's a lot better than snappy was able to do on my at-work dataset, and the majority of that dataset is HTML generated by Word…
[21:14:08] <mrmccrac> the larger one im trying now is 2783G, but its looking like at the rate its going now it will end up being around 2400G
[21:15:45] <Torkable> if I make a request but the connection has dropped or hasn't been set up yet, it just sits
[21:16:11] <GothAlice> mrmccrac: Compression works by building up a "dictionary" of patterns; the more data, the better the compression will be, generally. Thus the first half will take more space than the second half.
[21:17:05] <mrmccrac> yeah ill give it more time i guess, this is my 2nd attempt, i changed things so there will be larger documents to avoid too much document overhead
[21:18:22] <GothAlice> Torkable: connectTimeoutMS, socketTimeoutMS are both network-related timeouts you have control over at the MongoClient level, defaulting to 30 seconds.
[21:18:32] <GothAlice> Torkable: As seen: https://github.com/mongodb/node-mongodb-native/blob/2.0/lib/mongo_client.js#L178-L184
[21:20:21] <GothAlice> Torkable: As a fun note, https://github.com/mongodb/node-mongodb-native/blob/2.0/lib/mongo_client.js#L334 seems to indicate "no default" (zero/wait forever) for socketTimeoutMS. >:(
[21:24:30] <GothAlice> mrmccrac: I wish you luck with WT, BTW. I can see how for your dataset compression would be a very, very enticing feature, however in my own testing, and in the testing done by many others, there are some significant outstanding issues, some of which put data stored in WT at increased risk.
[21:32:07] <Torkable> problem is that it catches the failed connection and logs it, but server keeps running
[21:45:34] <Torkable> but how is there not a good way to deal with queries that don't have a connection