Log file Viewer
#mongodb logs for Friday the 15th of July, 2016
[00:57:46] <MacWinner> freem|nd, you should be able to use dotted notation.. like schema.index({'parent.sub.field': 1})
[01:14:39] <minnesotags> I've searched, but can't find the answer. I have used "mongod -dbpath" to ensure the directory is /var/lib/mongodb. I can see that the dbpath is /var/lib/mongodb, but if I run mongod --repair, or mongod -shutdown, it still looks for the dbpath of /data/db and says it can't find the server. How do I remove this reference to /data/db?
[01:16:25] <IronMike> I'm new to MongoDB and NoSQL. I'm building a web service and want to use MongoDB as the datastore. However, I'm concerned about the consistency model for things like security tokens and such. Is this outside the scope of use cases for a db like mongo?
[07:00:06] <Zelest> i do! why does all the media say "nice truck attack" when it's not nice at all? :(
[08:10:04] <Lonesoldier728> Hey so I am trying to call not equal and a second arguement not working though
[08:25:27] <koren> Lonesoldier728: can you show the document you want to query? Your request seems good. Also, do you use the mongo shell directly or are you using a driver?
[08:31:43] <koren> sorry I don't know mongoose, but your query is mongodb correct, so maybe this comes from your model?
[09:05:52] <obiwahn> Derick: is there some wiredtiger related development channel or is that now all mongodb internal?
[09:14:48] <Derick> obiwahn: best bet is probably the mongodb-dev google group: https://groups.google.com/forum/#!forum/mongodb-dev
[09:19:54] <obiwahn> At the moment I am measuring wiredtiger against rocksdb. It seems to be a problem when the data does not fit into wiredtiger's cache. I wonder if I miss some configuration option..
[09:22:11] <obiwahn> Does anyone happen to know if this is the "normal" behaviour? I got 300 * 10^6 key value pairs. And running over a selection of 30 * 10^6 elements takes about 6000 seconds:(
[09:24:31] <Zelest> obiwahn, yeah, but I can't wait for skylake support in BSD, so I settled with broadwell :)
[09:24:50] <Zelest> will probably regret it when skylake support is out, but meh, that's a problem for future Zelest :)
[13:47:04] <atbe> But, socketexception is raised immediately after. The address is available on my network and the port 27017 is open.
[13:50:19] <cheeser> could be network issues, yeah. it's not far after *this* code that we're down in to jdk/netty code.
[14:00:09] <atbe> When we began using the Mongo Java Driver, my colleague did not parse through the hosts provided to the MongoClient constructor to check if they were online in the first place (some of the host were down due to EngOps configuring things). It could be useful to instead of providing a 10000 timeout limit before raising SocketExceptions to see if the Socket is
[14:01:14] <atbe> I wrote a simple hostIsUp function that I would use to filter the potential hosts I would pass into the MongoClient constructor.
[14:01:14] <cheeser> the list of hosts is a seed list. the client will attempt to connect to the first it can find and discover the topology of your cluster.
[14:03:12] <atbe> I agree, it is a race condition but at least a warning in the logs would help so users do not produce hanging threads if they for some reason need multiple MongoClients to target different components of the cluster.
[14:03:35] <atbe> This way they can choose whether they want to create potentially unresponsive MongoClients
[14:16:17] <cheeser> you can always file a jira to that effect. that's going to be more effective than an irc conversation ;)
[14:50:42] <koren> sure! with the $ operator. db.col.update({'arrayItems.property': 'match me!'}, { $set: {'arrayItems.$.matched': true}})
[14:50:52] <saml> db.articles.update({author: 'Web Scale'}, {$set: {author: new Array where author Web Scale is replaced with WebScale
[14:51:37] <saml> Acts as a placeholder to update the first element that matches the query condition in an update.
[14:53:35] <saml> db.docs.update({author:'foo'}, {$set:{'author.$': 'Foo'}}, {upsert:false, multi:true})
[14:55:25] <koren> { author: ['foo', 'bar', 'foo'] } after your query will be { author: ['Foo', 'bar', 'foo'] }
[14:56:10] <koren> so it may not be suitable for your needs, for one-shot maintenance its ok but maybe you need something else
[15:00:50] <Lope> I'm running this update function update({},{$pull:{'profile.wish':webref,'profile.own':webref}},{multi:true});
[15:01:07] <Lope> it's unfortunately causing 6 updates, when only 2 documents have the webref in question.
[15:08:37] <Lope> yes. this works, but I'm not sure if it's ideal: {$or:[{'profile.own':'foo'},{'profile.wish':'foo'}]}
[15:12:54] <Lope> does it make any difference whether I search for an element inside an array in a document using $eq or $elemMatch?
[15:13:39] <koren> yes, elemMatch will return documents that matches all the conditions inside elemMatch
[15:14:29] <Lope> So $elemMatch is when you have multiple conditions, and $eq is simply one, equals?
[15:14:37] <koren> { elem: [{ a: 1, b: 2}, {a: 3, b: 4}] }, { elem: [{ a: 1: b: 1}, { a: 3, b: 4}] }
[15:15:39] <koren> with $and and $eq in the document I pasted you will have hard time to get documents that have element in array with a: 1 and b: 2
[15:16:24] <koren> because it will return this document too: { elem: [{ a: 1, b: 1}, { a: 2, b: 2}]}
[15:32:39] <Karbowiak> any of you able to help me figure out where i'm going wrong with the following query: http://pastebin.com/0g9pFa6j
[15:34:05] <Karbowiak> in sql it returns a list of regionIDs, together with the amount of kills it's found - but in mongo i just get an array like: array("_id" => array("source" => regionID), "count" => 2.0)
[15:40:47] <Karbowiak> hm, now it says: "errmsg" : "A pipeline stage specification object must contain exactly one field.",
[15:45:52] <Karbowiak> only question left really, is it possible to make _id become regionID? or should i just accept it :D
[15:49:35] <koren> I think you can also { $id: 0 } in the $project to remove it completely, otherwise pipeline keeps it
[16:03:46] <braqoon> Hi I'm trying to transfer db.collection.find().sort({$natural:1}).limit(1); to pymongo syntax. Got db.collection.find(sort=('$natural', 1), limit=1) but throws me an error "too many values to unpack". Any thoughts ?
[16:29:36] <btorch> like it's choosing "name" : "mongo1:27017" but instead I want to use ip or has to be somethine like mongo1.local
[16:30:13] <koren> you should put it in the hostname of the replicaset, ensure it is resolvable and you can use it when doing your rs.add
[16:35:34] <btorch> koren: not sure what you mean, cause I don't see any option for that to include in the mongod.conf under replication
[16:38:05] <btorch> and then tried rs.conf() but that just gives back "not authorized on admin to execute command
[17:42:54] <atbe> https://docs.mongodb.com/manual/reference/configuration-options/ I found this but it does not specify output that is not local
[18:21:30] <kurushiyama> btorch The name is supposed to be a DNS hostname, which mongo uses to resolve the IP.
[18:23:54] <kurushiyama> btorch And actually, you should never use an IP, not even an A record for replica set member addresses. Actually, the suggested way of adressing replica set members and config servers are CNAMEs. You should only deviate from this best practise unless you _really, really, really_ understand what you are doing.
[19:00:06] <theGleep> Hi, all - I have another question later ... but first; the welcome message says there's a searchable log of conversations? How do I search it?
[19:31:47] <theGleep> I did ... and I got: a calendar, a list of channels, one day of conversations. I couldn't find any way to search the history. :(
[19:34:16] <theGleep> I kinda figured that :) ... but is there a way to search across days? I'd like to see if my *real* question has been asked before I ask it again ... but it's not worth the time browsing through who-knows-how-many-days of log to find it
[19:35:17] <theGleep> Bummer .. pmxbot said "searchable" when I logged in, and I was hoping I could find it in the histories.
[19:36:13] <theGleep> So, here's my *real* question: Using Node.js and the promises version of the driver: mongo.connect(url).then(...) - what's the best way to handle closing the database?
[19:38:40] <koren> I personaly do something quite ugly but I register with the promise on my app start and add the db into global.db, then I access it everywhere using directly db., and I register a handler for sigint/sigterm to perform a db.close
[19:40:40] <theGleep> koren: I played around with that approach and bailed on it...don't remember why...I think it had to do with timeouts or disconnects or something...
[19:41:25] <theGleep> What I'm doing now is akin to the documentation (which only shows the callback approach, not the promises approach) - and has collection.db.close() buried *deep* in the callback chain.
[19:41:41] <koren> theGleep: yeah I understand. I don't handle the connection drop errors to make the service crash on purpose, then it is started again so I have not this kind of issue (a service running without connection)
[19:42:01] <koren> I read something about connection close, you should keep it open for the lifetime of the app and never close it
[19:43:25] <koren> I was talking about db close sorry, just talked about connection since it is actually a tcp connection
[19:44:24] <theGleep> The 3.2 docs seem to indicate the db should be closed after every call ... it looks like it's inherent in the architecture (based on the docs)
[19:46:40] <theGleep> You *can* get the DB in a then-able: MongoClient.connect(url).then(function(db) {})
[19:47:08] <Karbowiak> hey guys, me again. I've setup a text index on a string in a collection, the names can have weird symbols, including -, when i'm doing db.getCollection('corporations').find({$text: {$search: "4M-Corp"}}) it never actually matches 100% on 4M-Corp, instead it just matches everything, any way to make it take the - into account as a part of the string?
[19:48:28] <Karbowiak> if it's just 4M it finds 4 results as expected, but as soon as the - is there it just matches everything
[19:48:33] <theGleep> Big-K: I would guess that your search is using user input, so you can't easily convert it to a regular expression...
[19:50:36] <theGleep> that's what I was thinking - but it requires a "hard-coded" regular expression. But if the regular expression works, it wouldn't be that hard to create a text equivalent and do a replacement
[19:51:58] <theGleep> Here's a wild side-idea...maybe you could include a property "words" that would be text-field.split(" ") (or a regular expression to split on any whitespace) ... then you could search for a word...
[19:54:16] <theGleep> interesting ... that implies that the parser *honors* an exact-term spec; but by default assumes something else.
[19:54:37] <theGleep> I'm a noob, btw, so you can ignore anything I have to say ... unless it's a good idea :)
[19:54:44] <Karbowiak> db.getCollection('corporations').find({$text: {$search: "\"4M-\""}}, {score: {$meta: "textScore"}}).sort({"$score": -1}) neat, this will do
[19:57:05] <theGleep> the only way I can really see to make the code cleaner is like you said you did, koren: create an object that keeps the database around ... but the only way to account for accidental closures is to call "getDB" which would re-connect when needed...
[19:57:30] <theGleep> (No prblem Karbowiak - makes me feel better to give back when I'm looking for help myself! :)
[19:58:51] <theGleep> ... and if I have to call getDB() anyway...the only value I'm getting is to have the "db.close()" be centralized somewhere ... and that creates the other problem of how do I fire that other somewhere? I'm not aware of any inherent "onunload" event for objects during GC...
[20:00:49] <koren> well in our services we do a db connect, add db in global and let the service crash it db is not available, so it is restarted (docker container with auto restart)
[20:01:07] <theGleep> with promises, it seems there *could* be something like connect(url).then(db => {}).finally(db => db.close)
[20:01:47] <theGleep> Yeah .. I am figuring that at least one of my projects will need to have "forever app.js" on it. Don't like that approach, tho
[20:02:42] <theGleep> I'm actually looking into using the Bluebird promise library, which provides the concept of a "disposer" - but I haven't figured out how that works. The docs are written from an insider's view, so I'm having trouble really understanding how to use it.
[20:06:16] <Karbowiak> daym, it seems the php driver doesn't support the $meta operator, so i can't sort by score
[20:06:56] <Karbowiak> $find = $this->collection->find(array("\$text" => array("\$search" => "\"$searchTerm\""), "score" => array("\$meta" => "textScore")), array("sort" => array("\$score" => -1)))->toArray();
[20:14:09] <theGleep> So - for those checking in but not scrolling back; here's my question. What's the official best practice for db.close() when using the promises version of the API that's exampled here: http://mongodb.github.io/node-mongodb-native/2.1/tutorials/crud/
[21:37:18] <theGleep> Ah ... right. I realized that a few minutes ago -- and promptly forgot it again
[21:37:33] <JFlash> Failed global initialization: FileNotOpen Failed to open "/var/log/mongodb/mongod.log
[21:43:35] <JFlash> joao 21554 0.0 0.0 9496 2184 pts/1 S+ 18:32 0:00 grep --color=auto --exclude-dir=.bzr --exclude-dir=CVS
[21:45:01] <theGleep> Hmm...been a long time since I did a "kill" in *nix ... could you "kill mongod"?
[21:51:08] <koren> you should create a startup script in case your server restart, if you launch it manually
[21:51:32] <JFlash> koren, agreed, but as I see I'm running old ubuntu, this is going to be a never ending nightmare
[22:00:03] <theGleep> So - for those checking in but not scrolling back; here's my question. What's the official best practice for db.close() when using the promises version of the API that's exampled here: http://mongodb.github.io/node-mongodb-native/2.1/tutorials/crud/ ?