[01:18:37] <mdeboard> Hi. We are using mongodb replset (2 nodes, 1 arbiter) as our session store with Django. About an hour and some change ago, users reported they couldn't log in: session_db wasn't queryable for one reason or another. I did some digging and it appears that somehow the mongo.conf file had the "replSet = foo" line taken out.
[01:18:59] <mdeboard> This misconfigured node was the secondary.
[01:19:21] <mdeboard> So, human error there. But when I put "replSet = foo" back into the file and restarted, mongo threw an exception and went down
[01:19:59] <mdeboard> because of duplicate session keys, which must be unique. We had to basically restart the secondary, tell it it's not part of a replicaset anymore, so it woul dwipe its data.
[01:20:49] <mdeboard> Obvious human error of some kind with altering the config file, but why did Mongo behave in such a way? It seems like if the primary couldn't reach the secondary, tha'ts just a partition. But somehow this resulted in mutually inconsistent data, somehow?
[02:30:24] <sec^nd> What is the best way to query binary data?
[04:48:23] <Styles> I see the document store has a rough max size of 256gb per shard(unsure about this part). Does GridFS have any size limitations?
[05:35:25] <grahamsavage> is it possible to do analytical style (windowing funtions) in mongodb?
[12:45:19] <Industrial> weird idea, maybe: store all configuration for a REST resource as a document in the db and when the app server boots, get its configuration from the database instead of hard coded in a file.
[13:48:34] <bla> Why does creating an index (background: false) block reads from the table?
[13:49:48] <kali> because the lock implementation is n-readers-or-one-writer
[13:50:48] <bla> Ok, so there's no implementation for index-lock which in fact is not a table writer and could leave it open - that's not a technical impossibility just implementation deficiency
[21:06:02] <blocks> I'm stuck trying to make a copy of a database with auth enabled (a la http://stackoverflow.com/questions/17882052/mongodb-returns-unauthorized-on-copydatabase). Is there a workaround for this?
[21:50:09] <benhowdle89> Hey guys having a massive head scratching moment! Has anyone got any ideas on my problem here: http://stackoverflow.com/questions/20459248/mongoose-casting-array-to-string-when-using-set
[21:56:56] <blocks> @joannac thanks, i saw that.. but to get the whole DB i'd be copying reach object for each collection... it's boggling my mind there's not a simpler solution?
[21:57:14] <Styles> Mongodb wont start... it says starting/running procid blah then it isn't running and in the error logs it throws Permission denied: "/sys/dev/block/9:2/queue/read_ahead_kb"
[22:01:54] <joannac> blocks: or mongodump/mongorestore
[22:03:00] <blocks> @joannac I've also been getting "unauthorized" when runnning mongorestore even with the correct user/pw - i assumed it was part of the same issue
[22:12:29] <blocks> @joannac trying to test again now... definitely have the right credentials, but i get stuck even before trying to dump the latest data. mongodump from command line with the credentials gets me assertion: 11010 count fails:{ ok: 0.0, errmsg: "unauthorized" }
[22:13:49] <blocks> but just when it gets to mydb.system.users... mydb.system.indexes dumps ok...
[22:14:19] <joannac> what are the permissions for the user you're using?
[22:51:33] <blocks> @joannac thanks, i added that role and got the dump. -- now i'm just searching for the correct syntax to restore the DB into a DB of a different name (to make my copy without touching the original)
[22:59:07] <blocks> @joannac and now mongorestore gives me assertion: 18 { code: 18, ok: 0.0, errmsg: "auth fails" }
[23:29:32] <joannac> blocks: are you using user/pass for a user in the new db?