Log file Viewer
#mongodb logs for Friday the 23rd of May, 2014
[04:02:38] <Reflow> but i did db.collection.ensureIndex({ field : "text" }) and it does not optimize it
[04:24:02] <nevan> i created a script in ruby to extract db size of all db of a single mongo instance
[04:24:44] <nevan> and due to which mongo instance became unresponsive though system resources were empty
[04:26:25] <nevan> or if you could redirect me to internals of mongodb, i am not able to find anything deep in mongodocs
[04:31:08] <nevan> the mongo became unsresponsie after the script executed succesfully and remained so until i did not killed and started mongod process
[05:49:23] <fatih> what is the preferred way to do this? update with multi selects only for one selector
[06:21:16] <joannac> fatih: how do you know which 20 you need to update? do they have something else in common you can use?
[06:36:42] <jhonkola> Documents have 10 fields, I query with full document, there are 3 compound indexes all covering 2 fields
[06:43:18] <jhonkola> when calling explain() on long running example query it seems to find the indexes nicely
[06:56:50] <jhonkola> http://pastebin.com/rDn3e2gP contains currentOp() output of the same query I ran from shell that had taken 163 seconds when I ran currentOp()
[06:58:23] <jhonkola> could the reason be that sometimes mongod fails to find suitable index for query
[07:05:24] <jhonkola> lot of messages coming from aws SQS and then inserted to wait for further processing
[07:07:19] <jhonkola> there are enough queries that if index determination fails even occasionally it will be a problem
[07:12:47] <jhonkola> the problem that follows from long running queries is that the message writers eventually all block while waiting for query to finish
[07:16:27] <jhonkola> Also during the problem query there is lot of disk activity, looks like mongodb is scanning the entire collection
[07:34:14] <jhonkola> http://pastebin.com/kX0neisX there is currentOp() and explain for 2 longest running queries (sorry for needing to obfuscate some values, might make it harder to understand)
[07:44:10] <jhonkola> when looking at currentOp() when query is stuck and doing same query from shell explain reports reasonable values
[07:48:36] <jhonkola> it has lot of query + insert pairs, query to determine if doc exists, if not then insert
[07:49:21] <rspijker> if your explained query uses values for the indexed fields that already yield no results, they are going to be very fast
[07:49:53] <rspijker> if your ‘real’ queries use values for those fields that yield a resulting set of millions of documents, which you then have to narrow down by the unindexed fields, that’s going to be sloooow
[07:51:18] <rspijker> so, if you run an explain on the exact same query that is currently in currentOps it also runs really fast?
[07:53:02] <jhonkola> when I copied the query expression from currentOp() output and ran it on shell
[07:56:10] <jhonkola> Could the cause be that query planner fails occasionally, then the long running times would make sense as the collection is large
[07:57:15] <rspijker> what happens if you do a count on the 2 indexed fields for the specific values in that query?
[07:58:02] <rspijker> so, if your query was aaa:xxx, bbb:yyy, … ,time: timeVal, indetifier: idVal, … then do a count on {time:timeVal, identifier:idVal}
[07:59:35] <jhonkola> the load has increased which may just have brought underlying problem to light
[08:00:30] <jhonkola> I guess I'll next try to work around the problem by checking uniqueness of documents by querying only on identifier and time
[08:28:54] <monu> I'm facing an issue with my sharded clusters, the secondary node on one of my replicat set crashed due to some hardware issue.
[10:08:28] <sn0lan> Can anyone give me a hand with a mongo issue im having? Bit of a newbie with it.
[10:08:34] <Jadenn> i have a simple query, namely [ '$and' => [ [ 'location' => $location ], [ 'enabled' => 1 ], [ '$or' => [ [ '$or' => [ [ 'start' => [ '$lte' => time() ] ], [ 'start' => 0 ] ] ], [ '$or' => [ [ 'end' => [ '$gte' => time() ] ], [ 'end' => 0 ] ] ] ] ] ] ];
[10:08:52] <sn0lan> I have done a find({ 'email': 'test@test.com'}) and this returns a cursor for me.
[10:09:03] <Jadenn> now, 'end' => [ '$gte' => time() ] is false, but the query is still returning results
[12:31:04] <Veejay> Hi, we're thinking of upgrading from 2.4 to 2.6, I had a look at the changelog for 2.6 and I couldn't find any major changes that would break anything
[12:55:33] <hglattergotz> Anyone familiar with geo spatial search performance issues? Have a 2dsphere index on 900K recs (US and Canada) with GeoJson Point. Queries with $maxDistance greater than 100 miles are very slow. 200 miles takes over 20 seconds.
[16:46:44] <doug1> I'm trying to create an admin user with mongo. Where do I run the db.createUser() command? The mongos? The config server? The shards?
[16:48:19] <IndianaJoe> not sure. I'm not using sharding. On a replicaset, it needs to be the primary
[16:48:40] <doug1> IndianaJoe: I have a cluster set up. I created the key file... modified mongodb.conf on all instances and restarted. See errors about auth in logs now. I suppose that's expected. Went to mongos box, added user, it was accepted. Now when I try and login to mongo with mongo -u ddd -p ddd i get a socket exception error
[16:50:07] <doug1> don't need the logs... I got this :) rror: 13106 nextSafe(): { $err: "error creating initial database config information :: caused by :: socket exception [CONNECT_ERROR] for s1r1/mongo-shard-s1-n1.prod.slicetest.com:2701...", code: 11002 } at src/mongo/shell/db.js:1210
[16:52:12] <IndianaJoe> i would still look into the logs on the primary - it should give the reason for the early socket termination
[17:17:10] <Fetch> I've read some docs/slideshares about setting mongodb on AWS, but I need some conceptual help. I want to present a single DNS name (mongo.foo.bar) for my internal clients to access for a mongo replica set. If I setup mongo.foo.bar to point to an ELB in front of the replica set, is the client going to get the full seed list from the first server the ELB shuffles it off to? (that would be great) Is a better solution to have each replica set member create an A
[17:25:11] <Fetch> kali: what is a solution to have a single DNS name be effective for connections to a mongodb replica set, then?
[17:25:42] <doug1> oh yeah... reminds me... with two routers... what would I typically put in front? a load balancer or would round robin DNS do? is this stateless or not?
[17:28:37] <kali> Fetch: it's a very bad idea. the mongo client and server exchange adresses of replica set members to be ready to switch over or adresse a secondary.
[17:28:38] <Fetch> kali: alternatively, I run a zookeeper cluster, so I could have replica set members add themselves to a config space there, if there's any real client support for that
[17:30:33] <Fetch> kali: so, generally speaking, everyone keeps their mongo clients configured with a full array of replica set members?
[17:31:23] <kali> Fetch: more or less. but you don't need the array to be exact. as soon as the client manage to connect to one working member, it will pull the actual configuration from the replica set
[17:33:16] <kali> Fetch: you can imagine to have one dns alias that you use to configure the client, a CNAME pointing to one member. the client will connect and pull the full correct config when starting
[17:34:18] <kali> Fetch: but if your "special" node is down, a new client will fail to start. better to list all the replicas in the configuration string to the best of your knowledge, and let mongo adjust what's wrong
[17:34:29] <Fetch> kali: what would be the issue with having multiple entries for the same CNAME record? The client would pull one at randomly, connect, and pull a proper config
[17:36:31] <doug1> I'm trying to set up auth... where would I typically run the addUser() command? the shard? the config server? the mongos? the docs don't say
[17:48:28] <MTughan> I'm making a python program to help set up a sharded cluster and am trying to find a way to determine if a database and a collection has been sharded. I don't see any commands to do this in the Python API or in the database commands section, although the shell makes this available through sh.status().
[17:56:34] <doug1> so... I added the admin user on mongos box. Seems to work... also seems to have 'replicate' to config server.... but the shards I can still mongo to without a password...
[17:57:20] <MTughan> I don't actually know how users are supposed to propagate, so maybe I shouldn't say anything…
[17:58:16] <doug1> I tried to connect from the config server box to the shard box, and auth denied. Ok... so I did it again passing in user/pass and it fails. Huh?
[17:59:25] <doug1> works tho from shard to config box... so obviously the user isn't being replicated to shards (but it is to config server)
[18:23:47] <MTughan> I'm making a python program to help set up a sharded cluster and am trying to find a way to determine if a database and a collection has been sharded. I don't see any commands to do this in the Python API or in the database commands section, although the shell makes this available through sh.status(). Is it possible to get that from Python without querying the config database?
[18:56:44] <jbrechtel> Is there a 'right way' to create mongo users using the Ruby mongo driver? I'm finding db.eval() to be a painful way to do it. In particular, I'm unable to create a second user after I create an admin because I get an 'unauthorized' error.
[18:57:19] <jbrechtel> I've tried authenticating as the admin after I do that but then I still get unauthorized. Am I perhaps racing with the create action by executing the authentication immediately following the create?
[18:57:44] <IndianaJoe> once you have created the first[admin] user, you'll need a db.auth() before you can create others
[18:57:56] <jbrechtel> To be clear, when I try to authenticate as an admin after creating the admin I get unauthorized on the db.auth() for the admin user I just created
[18:59:10] <jbrechtel> hmm.... I'm using these: ["userAdminAnyDatabase","readWriteAnyDatabase","clusterAdmin"] ...and mongo 2.4
[19:29:12] <jbrechtel> Contrary to what I said earlier, authenticate does not fail. authenticate DOES fail if I try db.eval(db.auth()) but db.authenticate() does not fail. However, even with db.authenticate() the subsequent adduser fails.
[19:41:40] <jbrechtel> I don't know why I care about creating that user in the admin database. The annoying thing seems to be trying to authenticate as an 'any database' admin with the Ruby driver. The authentication, for whatever reason, doesn't seem to carry to operations across databases. It's not the same as doing a 'use someotherdb' after authenticating in the mongo console.
[20:14:16] <ekristen> I’m trying to dump documents from a collection where a date is great then and less then two dates
[20:14:38] <ekristen> but when I go to run mongodump, it wants to dump the entire contents of the collection
[22:46:32] <PXER> Mac OS Mongo installation of 2.6: Every executable is around the same size. What gives with that?