Log file Viewer

[00:00:31] <pEYEd> cheeser: thank you!

[00:00:58] <cheeser> getCollection() rather

[01:50:03] <Casan> question: are configuration and query router servers needed if sharding is not used?

[01:50:31] <cheeser> mongos and configsvr ? no.

[01:50:43] <Casan> question2: will configuration and query servers have any use if a mongo cluster is only using replica sets?

[01:50:54] <cheeser> mongos and configsvr ? no.

[01:51:28] <Casan> thanks, so for a good normal setup, 3 mongods with replSets and 1 arbiter is good, correct ?

[04:36:26] <Creeves> hi guys, this should be a relatively simple question I think. I'm attempting to update an array of objects and have the index of the element I want to update in a variable

[04:36:44] <Creeves> i'm just unsure of the syntax to specify that index as the array element i want to set

[04:36:54] <Creeves> i"ve made a gist here: https://gist.github.com/Creeves/98f78901b35e25b88a48 - thanks again :)

[04:38:39] <Creeves> I've read that something like: " $set: { index_details.3.status: ... " may work, but am unsure how to use the variable holding the index rather than it being hard coded

[04:42:45] <Creeves> ahh I think I worked it out, I need to build the key used as a string first

[12:54:45] <janemba> hi

[12:57:19] <sum1else> Hey guys. I have a sharded replica set. 3 primaries each with 2 secondaries and 3 config servers. I had turned auth on and was unable to log in so I turned it off again. Now I am noticing that data is no longer replicated between, at least, the primaries, as far as I can tell

[12:57:31] <sum1else> Does anyone know how I can troubleshoot this?

[12:57:58] <sum1else> Also, after I turned auth off, the users in admin was empty

[13:03:49] <synapse> Guys when I run the shell and it says connecting to test, where is that db? the test one?

[13:07:04] <sum1else> synapse, look for dbpath in mongod.conf and it will be in that folder

[13:09:33] <Fernandos> hi

[13:10:07] <Fernandos> I added a new array field to the scheme. How can I update old objects missing that field?

[13:10:42] <StephenLynx> update({field:{$exists:false}},{field:[]},{multi:true})

[13:10:50] <StephenLynx> or just handle that in application code.

[13:10:56] <Fernandos> Or shouldn't I explicitely update, but instead just do this onEdit

[13:11:05] <StephenLynx> onEdit?

[13:11:26] <Fernandos> I mean on update

[13:11:38] <Fernandos> when someone edits a form

[13:11:50] <StephenLynx> usually I just do object.field = object.field || [] so I don't have to update the database.

[13:13:58] <Fernandos> hmm would it be wrong if you do object.field = [] instead?

[13:14:01] <Fernandos> that's what I had

[13:14:38] <StephenLynx> that would overwrite if the field actually exists.

[13:15:58] <Fernandos> oh.. this is awesome.. I think it worked! simple but smart way of solving it =)

[13:16:43] <Fernandos> yes it works! wow

[13:16:55] <StephenLynx> it just werks :v

[13:18:02] <synapse> sum1else can't thank you enough

[13:19:03] <synapse> sum1else The weird thing is when i run db from shell it says it's called test. when I run show dbs it says it's called local.0. In the dbpath like you said, I see the db called local.0 I just wonder where it gets the name test from

[13:20:15] <StephenLynx> test is the default db

[14:24:37] <dcorbin_> We use mongo 2.6. Is there a reason I should use ensureIndex() vs. createIndex()?

[14:25:48] <dcorbin_> I understand that ensureIndex is deprecated in 3.0

[14:28:51] <Doyle> Hey. How does one recover from: [initandlisten] exception in initAndListen: 17329 no NamespaceDetails for index:

[14:30:26] <Doyle> That's what I'm getting when trying to restart mongod. When it failed: "63e48d6df5.actions_custom Invariant failure nsi.details( descriptor->indexNamespace() ) == NULL src/mongo/db/catalog/index_catalog.cpp 442

[14:30:26] <Doyle> " and a few lines further "***aborting after invariant() failure"

[15:22:20] <sum1else> Hey guys. I am creating a cluster but once I move on to try and set up authentication, things just seem to stop working properly. I am getting locked out of the service or I am seeing replication and/or sharding not working

[15:22:26] <sum1else> can anyone help out?

[15:27:34] <alexbevi> not sure if this is an "easy" question, but can libmongoclient be build from source in 3.x?

[15:28:03] <alexbevi> in 2.x you could do this with `scons mongoclient`

[15:36:23] <sum1else> alexbevi, "An obsolete library. Don't use it in new projects. http://algernon.github.io/libmongo-client/" - https://github.com/algernon/libmongo-client

[15:39:38] <alexbevi> @sumlelse: I was building this right from the mongo source though (see https://github.com/mongodb/mongo/blob/05bebf9ab15511a71bfbded684bb226014c0a553/SConstruct)

[15:40:03] <alexbevi> is that the same thing?

[15:43:52] <alexbevi> i'm trying to upgrade robomongo from 2.4 to 3.0.6

[15:44:07] <alexbevi> and it includes the full mongo source, and the mongoclient was being built from that

[15:46:07] <sum1else> I don't know

[15:59:49] <alexbevi> sum1else: thanks for responding :)

[16:00:49] <sum1else> Sure. If I knew more I'd toss my 2 cents in

[16:12:57] <Zelest> Reading up a bit on locks in db.currentOp() .. what exactly does "Shared Lock", "Exclusive Lock", "Intent Shared Lock" and "Intent Exclusive Lock" mean?

[16:13:07] <Zelest> Ref: http://docs.mongodb.org/v3.0/reference/method/db.currentOp/#currentOp.locks

[17:06:06] <beekin> Curious, will I actually see faster speeds using MongoDB for a basic blog rather than something like postgres?

[17:06:43] <beekin> Or is the performance benefit only really noticeable when complex joins are being used and MongoDB comes to the rescue with embedded documents?

[17:43:58] <cheeser> beekin: i used to work for a company offering blogs backed by mongo. we did millions of hits a day.

[17:45:29] <beekin> cheeser: I have no doubt that Mongo will be suitable the people I need to convince like to talk about speed.

[17:45:53] <beekin> sorry, *but the people

[18:38:57] <preyalone> Is this still true in Mongo v3? "In fact, MongoDB throws the requests on every indexes independently (and also throws an index-free request), and awaits for the fastest to achieve, then terminates the others." http://blog.quarkslab.com/mongodb-vs-elasticsearch-the-quest-of-the-holy-performances.html

[18:40:18] <cheeser> how old is that post?

[18:40:33] <preyalone> march 2015

[18:40:35] <cheeser> the query planner was rewritten in 3.0

[18:40:54] <cheeser> so, probably, yeah. though I think that's only true if there isn't a cached query plan

[18:55:15] <jr3> What is the best way to push database changes to a prod server

[18:55:29] <jr3> lets say I need to add an index to a huge collection

[18:55:39] <jr3> do you just run that on prod willy nilly

[18:56:08] <beekin> Through the shell?

[19:06:08] <yopp> https://jira.mongodb.org/browse/SERVER-20150

[19:06:10] <yopp> fuck

[19:06:15] <yopp> this is insane

[19:07:52] <yopp> So with 3.0 we just can't use shard balancer, because it will block map/reduce for indefinite time

[19:12:10] <cheeser> well, one will block the other for the reasons stated.

[19:16:11] <joshua> If you had a good shard key wouldn't the balancer not be running all the time... unless you horizontally scale a lot

[19:18:10] <n0arch> on a member of a replica set, is `kill SIGUSR1 $pid` the preferred method for log rotation over `cp log{,.bak}; > logfile`

[19:18:30] <n0arch> ie just echo'ing into the log

[19:28:02] <yopp> cheeser, for _hours_?

[19:28:30] <yopp> I understand that for integrity you can't live without the locks, but this is bullshit

[19:29:01] <yopp> Basically Shard Balancer can break your app for indefinite and unpredictable amount of time

[19:31:45] <yopp> For some reason I thought that thing is: in the background, when there need for a balancing, one shard is copying chunk to another shard in background. Then at some point shard locks for a period of time required to accept copied chunk on recipient node, and to tell originating node that chunk is accepts.

[19:32:04] <yopp> And if chunk changed in the meantime, then operation aborted.

[19:32:15] <yopp> And repeated

[19:32:54] <yopp> Because locking the whole collection is fine, but only for short period of time (seconds in the worst case)

[19:52:30] <je> Would the Java MongoClient behave any different if it was instantiated using a ServerAddress vs. List<ServerAddres> containing only one ServerAddress?

[19:52:56] <beekin> Besides impedance mismatch - are there any other advantages to using mongodb for small scale sites?

[19:54:32] <cheeser> je: yes

[19:54:59] <cheeser> with a List it will discover the topology and try to talk to the primary even if it was another machine. with the single ServerAddress it will only talk to that one server.

[19:55:08] <cheeser> beekin: what impendence mismatch?

[19:56:18] <beekin> cheeser: Sorry, the lack of impendance mismatch being the main advantage. I realize it's "cool" to work with objects and JS like querying but I can't really identify any other benefits for small sites.

[19:57:13] <je> cheeser: thanks for the explanation

[19:59:14] <beekin> It seems that for smalls scale sites...it comes down to flavor preference.

[20:10:56] <cheeser> je: any time