Log file Viewer

[11:56:41] <suchkultur> We just upgraded from 3.4.9 to 3.6.3 and my impression is that our CSRS shows a “chicken-and-egg problem” now. We issued the `db.adminCommand({setFeatureCompatibilityVersion: “3.6”})` command and it returned with this error message: `{

[11:56:43] <suchkultur> "ok" : 0.0,

[11:56:45] <suchkultur> "errmsg" : "expected entry { _id: \”<redacted>\”, lastmodEpoch: ObjectId('59f334ee46b31acfe81b6d7c'), lastmod: new Date(4294967476), dropped: false, key: { _id: \"hashed\" }, unique: false } in config.collections for <redacted> to have a UUID, but it did not",

[11:56:47] <suchkultur> "code" : 1,

[11:56:49] <suchkultur> "codeName" : "InternalError"

[11:56:51] <suchkultur> }

[11:56:53] <suchkultur> `. The result is a partially upgraded or downgraded state. What i don’t understand is, that UUIDs are a 3.6 feature, so why does stepping up from 3.4 to 3.6 leads to a problem with non-existent UUIDs on collection level (tested with db.getCollectionInfos())? And what are the implications of this situation / how to proceed?

[12:47:14] <bitknight> Can anyone tell me if using a view in the from field during $lookup has poor performance?

[12:48:02] <bitknight> I am using an aggregate query on a collection with 765 documents. One stage is performing a $lookup against a view with ~2600 documents. The foreign and local fields are both indexed in their respective collections. The runtime I'm seeing for this is ~12s. Replacing the view in the from field with the source collection it was created from (with ~20k records) finishes in 0.05s.

[14:09:35] <suchkultur> Regarding my question: The problem was apparently, that i did not issue the db.adminCommand({setFeatureCompatibilityVersion: “3.6”}) through a mongoS node but on the primary of one of the shards.

[14:10:16] <suchkultur> I connected to a mongoS and issued the command again. The issue seems to be resolved now.

[14:39:04] <Omnipotent> Hi If one has few millions of records in Mongo and they want to distinct on it should they do Mongo distinct or fetch all the records matching a query and then do distinct at application level?

[14:39:07] <Omnipotent> Side note assuming the find query in both the cases is going to be indexed, that is when doing distinct or when doing a find the query part will be indexed.

[16:11:22] <mgaunard> hi, I've got a field which can either be a double or an int32

[16:11:36] <mgaunard> I'm using the C++ API and I'd like to just extract something convertible to double

[16:11:43] <mgaunard> get_double requires it being a double exactly

[16:11:49] <mgaunard> what's the best way to achieve this?

[22:31:07] <sphet> Hello, I have a collection of documents with 4 keys: world, level, zone, sector - I would like to aggregate the results so that the item is an array of world, with each world containing an array of levels, where each level contains it's zones and each zone contains it's sectors.

[22:31:39] <sphet> I wonder if this is possible with the aggregate framework. I have been working with multiple groups in the pipeline but I seem to be developing too many nested arrays or keys

[22:41:30] <sphet> Is it simply just nesting grouping id's and pushing?