Log file Viewer
#mongodb logs for Tuesday the 21st of August, 2012
[00:33:03] <addisonj> just curious how performance was for high loads, I am pretty far into going AWS, mostly for more scalable mongo, but was curious what others thought
[00:34:20] <addisonj> it seems like their higher end instances do offer some decent IO, and they have some fancy management stuff for repl sets
[00:39:05] <TkTech> addisonj: We currently run on AWS EC2 and are happy with it. Within the same region the IO bottleneck is pretty much non-existent with our fairly high update and insert load.
[00:39:35] <TkTech> addisonj: What ends up killing us is the map reduces on 2.0.6 which for whatever reason chuck cannot yield warnings every couple thousands record iterations
[00:40:13] <TkTech> addisonj: Yes and no, the primary runs on Amazon's new high-up instance with 60-something GB of RAM.
[00:40:52] <TkTech> Our dataset is many times our available RAM, and page faults were becoming a bottleneck
[00:42:46] <addisonj> thats the other thing about EC2, so much flexibility in scaling up, joyent/rackspace/everyone else don't give that, but just the configuration/management overhead is making me feel different atm
[00:43:36] <TkTech> Well, most of the MongoDB cloud providers are just reselling AWS and OpenStack providers.
[07:39:35] <stabilo> so. regarding pymongo... let the following statement be a query that is valid in the mongo console
[07:41:02] <stabilo> Now, how do I construct the very same query from within python? I assumed it has to be
[07:43:43] <fleetfox> hey guys. { $ne : true } will return true for false and where key does not exist?
[07:46:09] <stabilo> for the mongo console it's 09:39 < stabilo> > db.fs.files.find( { filename : /^8/ } )
[08:08:16] <_johnny> stabilo: yeah, i had a "run in" with python regex aswel. besides what NodeX, if you want to do a { field: /^something/i } (case insensitive), you can use { "field": re.compile("^something", re.IGNORECASE) }
[08:15:57] <stabilo> _johnny: looks at least cleaner, but the re.compile() should take more time than just defining regex = '^8' (I want to match files that have a filename starting with 8)
[08:20:03] <_johnny> stabilo: how so? it is my understanding that a compiled regex obj is more efficient. at least when used multiple times. i'm not aware how pymongo translates it however
[08:42:51] <neil__g> i have a P-S-A replica set configured. When I do batch imports, the secondary's replication delay just increases and increases until it's an hour behind the master. Is there any way to force a sync or give replication priority?
[08:48:16] <Bartzy> I update a document and using upsert - if the document was only updated, can I get the _id without doing find again ?
[09:20:28] <NodeX> then when I run my analytics I parse out what the users have been doing and archive my history collection
[09:20:51] <NodeX> I typically parse 500k documents at a time and then export them elsewhere incase I need the data for somehting else later
[09:22:16] <justin_j> I'm processing the data as it lands attempting to place it into aggregate buckets and so on
[09:24:19] <justin_j> current pipeline is that I receive events from apps, queue them, and then the aggregator reads from the queue
[09:27:17] <NodeX> if I was doing it on a minute basis I would esitmate traffic per minute and limit by that
[09:28:09] <NodeX> the good thing about doing it that way is there is already an index on _id so you dont need anymore :)
[09:29:53] <justin_j> the latest issue I've got is that I've used a doc structure for storing unique user counts (and other counts) and I'm trying to allow filtering access by attributes (color=blue, item=shirt) etc
[09:32:26] <justin_j> allowing filtering on attributes is turning out to be difficult - I thought I had a working solution, but it's flawed
[09:34:21] <justin_j> so, I store something like: { ',color=blue,item=shirt,': { c: 645434, u: 1232 }, 'color=red,item=shirt': {c: 64324, u: 343 } }
[09:35:41] <justin_j> then, for a given day I find the doc and iterate through the fields. My idea was that I can do a string match on attributes to filter.
[09:36:43] <justin_j> Now, you can imagine that a user would have a blue shirt in the morning and then a red shirt in the afternoon so his unique count would be present in both 'u's in the doc
[09:37:21] <justin_j> if I filter wholly on ',color=blue,item=shirt,' or ',color=red,item=shirt,' then everything is fine - he's only counted once
[09:38:39] <NodeX> out of interest why dont you store the item and colr in the attribs (with the c and u)?
[09:41:08] <justin_j> could do, so something like { c: 43235, u: 5432, attiribs: {color: 'blue', item='shirt'} }
[09:42:13] <justin_j> that would prevent 'c' and 'u' being used as attributes - those attributes are provided by the client
[09:46:47] <justin_j> for a given day, I get the doc and iterate through the fields matching on color and/or item, summing the 'u' and 'c' fields
[09:49:33] <justin_j> so for: { ',color=blue,item=shirt,': { c: 645434, u: 1232 }, 'color=red,item=shirt': {c: 64324, u: 343 } }
[09:50:11] <NodeX> in that case I would distinct the color=.. part and add the totals ( I guess that's what you're doing)
[09:50:26] <justin_j> if I specify I want the unique users for blue shirts I just iterate through looking for attributes color=blue, item=shirt
[09:51:16] <NodeX> if you store the color and the item as I suggested you can query in any fashoin you like
[09:52:02] <justin_j> which means they are present in both 'u' counts - if I just look for 'item=shirt' then it will count them twice
[09:53:25] <NodeX> in your itteeration do you know what users have been counted already at any point?
[09:53:50] <_johnny> i haven't read your entire discussion, but would union - intersect not give you the right number?
[09:54:47] <_johnny> ah, okay, sorry for meddling. i didn't quite understand your "color=red,item=shirt" schema :)
[09:55:08] <justin_j> so, my first solution was to say, 'ok, he'll be counted twice if we're just looking for shirts - can we put an offset somewhere to fix that?'
[09:59:10] <justin_j> the other thing I tried was storing and updating all combinations of attributes given
[10:08:14] <SpanishMagician> I need a database, there are users, users have clients and you make call and events for those clients
[10:10:03] <SpanishMagician> not programming events, event as in a show, a presentation, a meeting...
[10:13:27] <justin_j> it depends upon how many calls and events you can have per client - you could do it like that
[10:14:07] <SpanishMagician> so what is better separate collections or one collection for clients calls and events?
[10:17:16] <SpanishMagician> well, I don't think there will be too many users and information, it's a small app
[10:25:02] <SpanishMagician> let's assume something simple, users, users have clients and make calls and appointments to those clients
[10:25:44] <NodeX> I would have a users collection and a clients collection and an appointments collection
[10:27:48] <SpanishMagician> no, user calls client and write the results and maybe make an appointment
[10:30:20] <SpanishMagician> call is you call the client, maybe he's not interested or unreachable, appointment is you go visit the client
[10:30:34] <NodeX> it's your app, not mine, I would probably change my "appointments" colleciton into an "actvities" collection and do it all in there
[10:31:30] <NodeX> it's all down to your app and how you query the data as to the most efficient way to do things
[10:32:04] <SpanishMagician> best approach is have a collection for users, clients, calls and appointments, right?
[10:32:15] <Cubud> If you want to show an individual call on screen then I'd go for a separate collection
[10:34:25] <Cubud> Order+lines are a single entity, you won't need to have 2 users updating 2 different lines at the same time
[10:35:12] <Cubud> If you don't want 2 people making an appointment for a client at the same time, but want people to be able to edit a client while someone else raises an appointment then perhaps you need a ClientAppointments collection
[10:35:56] <Cubud> If you want to be able to make simultaneous changes to the appointments for a single client then have a collection where each appointment is its own document
[10:36:51] <Cubud> But I know nothing of MongoDB yet, these are just standard programming guidelines
[10:36:52] <NodeX> if however you are going to query that alot it can be expensive so it may make sense to have hot data embedded
[10:39:35] <elio_> and an array of a nested document which holds the author and content of the message
[10:45:53] <Cubud> Like a script of things to do in order to replay a conversation (or a subsection of it based on a time period)
[10:46:56] <Cubud> Just like on twitter, except the client only gets what is destined for that user based on username or group memberships
[10:48:20] <Cubud> When posting the message to a group you grab the group, see who is in it, then send it to those people
[10:49:41] <elio_> NodeX, it has nothing in special I want to know how experienced mongo db users model this problem on the database
[10:52:00] <NodeX> http://www.mongodb.org/display/DOCS/Retrieving+a+Subset+of+Fields#RetrievingaSubsetofFields-RetrievingaSubrangeofArrayElements <---- will get you last 50 elements of your messages
[10:52:39] <Cubud> I would have "Message" Sender, Recipients, GroupChatName (for memorandum purposes), Body
[10:53:20] <Cubud> When sending to a group you would grab the group, see who is in it, then put those people in the recipients list
[11:01:43] <Cubud> It seems that fetching data in MongoDB is slower as the record you seek is near the end of the data
[11:03:07] <NodeX> caching is maintained by the operating system, if you insert a record it -may- still be in memory and it -may- not, if you access that documnent a few times it will be in MRU
[11:03:40] <Cubud> But when I FindOne with EmailAddress = "1@home.com" it is quicker than when I look for 5000000@home.com
[11:03:54] <_johnny> NodeX: would one way to force it into MRU be to use the "return inserted document" when adding to the collection?
[11:06:46] <NodeX> I have a 140m document colleciton and I can skip/ limit and grab 1000 records deep inside the collection in well under a second
[11:10:16] <_johnny> NodeX: speaking of which, are there any "best practice" on indexing for regex purpose? or is that not possible at all?
[11:12:25] <_johnny> i mean, from what i can tell from the docs, indexes only work when you do regex queries, if you start them with ^, and don't use case insentive
[11:14:45] <NodeX> I currently use it for a 28million doc collection (UK Postcodes) with a prefix and can find (using a suggestive ajax style approach) a match in 1 few ms
[11:15:55] <_johnny> funny, i do something a bit similar. i have 2 mio address records, where i do ajax search complete for street names. i use a street name collection aswel to look up the street names first, to work around the case insensitive
[11:16:18] <_johnny> i was thinking of auto titlecasing it, which would work for most cases, but not all :(
[11:17:40] <_johnny> i'll just add that to the lookup collection. it's still relatively small, and the mongodb isn't in production yet
[11:19:14] <_johnny> how long did it take you to fill those postcodes in? :) i had an excessive xml schema which i had to parse when inserting to mongo (new schema). took 5 hours for 2 mio elements. there's something i'd rather not do again, lol
[11:19:14] <NodeX> I did it a couple years ago when I delved into the viability of mongo for a job board and wrote one as POC
[11:19:58] <NodeX> I have the postcode data that I have been building for years in Json that I exported from SQL, it took about 2 hours to insert and a few to index
[11:21:02] <_johnny> while i was loading the data in, i ran a periodical query to check places near a certain point. was fun to see how the "world" got populated as it went along :p
[11:21:12] <NodeX> with a 2d index so I can get the closest "large place" (town/ city/landmark) to a given lat/long
[11:22:03] <Cubud> I have to go out for an hour or so now, but here is my really simple test app source if you want to see what code is being executed that gets progressively slower
[11:22:27] <_johnny> i was thinking of running a map/reduce gimmick to find the most crowded places in the most crowded cities, and vice versa the most stranded :)
[11:22:46] <NodeX> I use it as a reverse Geocode for my apps as they all rely on lat/long and geo hashes
[11:23:23] <NodeX> MongoServer.Create("mongodb://localhost/?safe=true"); <---- can you not use a unix socket in your app ?
[11:25:06] <_johnny> http://www.mongodb.org/display/DOCS/CSharp+Driver+Tutorial#CSharpDriverTutorial-Connectionstrings doesn't look like it, no
[11:27:27] <_johnny> Cubud: i'm not that into it, but to me it looks like you're iterating all, rather than skipping?
[11:30:26] <_johnny> NodeX: a specific reason to use upper case rather than lower case, or just preference?
[11:33:20] <_johnny> ok. is there a upper/lower in mongo internally? like for .update({}, {$set: { new_field: old_field.toUpperCase() } }?
[11:45:02] <stabilo> I have 10 items in the gridfs, all having filenames starting with 8. I want to query only one. So I do
[11:45:41] <stabilo> which should limit the number of results to 1 for my understanding. Apparently it does not what I want since I get all 10 files
[11:47:15] <stabilo> findOne is not preferred since I want to get an GridOut object rather than a dictionary
[11:52:13] <stabilo> thank you. This gives me enough motivation to really start hating the pymongo doc.
[11:52:14] <_johnny> algernon: you're right. the this object in a cli update is the entire shell. hehe :(
[12:31:06] <Bartzy|work> I update a document and using upsert - if the document was only updated, can I get the _id without doing find again ?
[12:34:04] <algernon> on the wire: no, you can't. your driver may do it for you transparently though.
[12:47:57] <_johnny> phpmoadmin might be what you're looking for. personally i've used rockmongo a bit, but not extensively
[12:51:03] <stabilo> hrm. what is the best way to get a gridfile using a query object. I first thought that the collection.find() method will do it, but it returns only a cursor (over dictionaries) and I don't know to get a gridfile from that. One solution would be to extract the exact filename from the dictionary and make a second call to the database like db.fs.files.get_version(filename). But that sucks (TWO calls to the DB for ONE file)
[13:10:48] <zykes-> how can I do "relation" ish things as in saying what category a a product belongs to ?
[13:40:48] <circlicious> from reduce i would like to return all fields of the # documents from the collection i am working with - does that make ssense?
[13:57:01] <circlicious> yes, actually i want on a specific item_id, but for now i am doing all, as i think adding the item_id =x then will be easy anyway
[14:00:11] <circlicious> ok, forget that. there i meant i need to get documents with same created_at on item_id = x
[14:00:32] <circlicious> but right now i am doing on entire collection, will add item_id=x filtering later when i know how to do it anyway
[14:02:30] <circlicious> 1 thing i can now do is, map reduce can give me the created_at that appear 2 or more times. i can then query once for each created_at then
[14:03:33] <circlicious> i think my data will also ecome big soon ;D and i have no idea how i will handle it, but i'll try and manage
[14:46:42] <Cubud_> _johnny: Yes, I am doing 5 million individual selects. I am trying to see how quickly I will be able to serve a DB request. For inserting it is 20,000 per second (wow)
[14:47:17] <Cubud_> It's as if MongoDB is sequentially scanning the rows rather than using the index, because it takes longer to select data near the end
[14:57:20] <estebistec> In explain-plan output, is allPlans indicating indexes that were actually used in the query or simply considered by mongo because they all have related fields?
[14:58:20] <NodeX> the disk will be thrashing all over the place looking for the data as no MRU / LRU will be touched
[15:12:28] <_johnny> NodeX: i'm about to parse all my xml again, wanting to put it in a mongoimport complaint format. i'm trying to store it as json, but even for smaller files (36kb) i catch the exception BSON representation of supplied JSON is too large. i've added \n's as it looked like mongoimport reads lines at a time, but still same garble. am i doing something completely upsides down ?
[15:33:12] <Cubud_> Okay, so as I select index.ToString() + "@MyDomain.com" it finds them quickly when index is low
[15:53:28] <therealkoopa> If I want to provide users with the ability to reset their password, I'm doing it by generate reset tokens that are used. Is it better in mongo to create a separate small collection with pending reset tokens, that reference the user. Or to store the reset token information on the users collection?
[15:57:00] <therealkoopa> Cubud: I was worried about searching for the reset key if the users collection is massive.
[15:58:20] <estebistec> asking again, briefly as I can: in explain output, are all the indexes in allPlans actually being used or are they candidate plans?
[15:58:54] <Cubud> but you won't guess a key + emailaddress, especially if it has an expiry of 30 minutes
[16:05:10] <therealkoopa> NodeX: Well right now about 3. Hopefully when we deploy, it's going to be massive :). I'm a little worried about searching by reset key. I'll probably somehow encode the email address in the key, and search on that? Um
[16:06:06] <NodeX> I would (for sanity) use a front facing cache and not mongo for the initial lookup
[16:08:01] <therealkoopa> Ah, yea. So the reset key will be in redis and the value will be the user. Then I can look up the user quickly in mongo.
[16:08:59] <Cubud> make the link http://mysite.com/Account/ResetPassword?User=cubud&ResetKey=123456789
[16:09:42] <therealkoopa> What if the reset key is just the base64 encoded username? Too easy to guess?
[16:10:46] <NodeX> that way someone would have to hijack the session, guess the user agent, email - can add IP for good measure if you like too
[16:11:17] <Cubud> user + random number + a timestamp of when it was generated, so it can only be reset within X minutes
[16:11:42] <Cubud> So you have to guess the random number + find a user that has requested a reset + within X minutes of them doing so
[16:13:14] <NodeX> err salting the session id plus the user agent is far more secure than not salting it
[16:13:33] <Cubud> but with sniffing you don't need to understand the data, you just need to possess it
[16:14:02] <Cubud> The resulting code would be as meaningless as a random number, but still as simple to use to access the account
[16:14:05] <therealkoopa> Would you worry about them using a different browser for the reset? That's just a thought.
[16:15:30] <NodeX> so sending your "sniffed" data means nothing if you can't send me what my app salted in the first place to compare it against
[16:15:57] <jjanovich> question, my server is set in EDT timezone, but looks like mongo is returning in a different timezone
[16:16:34] <Cubud> the server sends an email with some key in it, if that key is sufficient to get the user in then it would be sufficient to get in the person that sniffed the key wouldn't it?
[16:18:12] <therealkoopa> Okay, so it's either store it as key/value in redis with an expiration, whcih is nice and easy. Or base64 encode some magic and have to deal with the timeout manually. THink they are both about the same on a 'goodness' level?
[16:19:21] <Cubud> NodeX, using the email address as my key gives me 12,500 lookups per second without any speed degradation
[16:21:09] <jjanovich> does mongodb store all datetime as UTC and I have to convert or is there a way to force it to same in local timezone
[16:26:38] <Cubud> I think that seeing the high speed of selecting against a specific key is enough to convince me to use Mongo :)
[16:27:41] <therealkoopa> Cubud: You don't have to worry about storing reset tokens on the user if throwing redis in front.
[16:44:15] <tsally> if I write a new document to a sharded collection, how do i know whena read of this document is guarenteed to succeed?
[16:44:57] <_johnny> ok, so i have a question about mongoimport. how on earth do anyone manage to import anything? i have a 20mb array file which is too large. any tools for splitting this seemingly "huge" file?
[17:01:55] <NodeX> can you do db.your_collection.find({Email:"some@email"}).skip(0).limit(1).explain();
[17:05:16] <ramsey> Derick: does that new beta driver also fix this issue? https://jira.mongodb.org/browse/PHP-377
[17:05:50] <NodeX> the only thing I can say is that your data is non sequential on disk and the reads to get the data are thrashing the disk
[17:06:25] <Cubud> I would say that is correct, because it is 1@MyDomain.com through to 1000000@MyDomain.com
[17:12:54] <ramsey> Derick: nevermind... we can't really test the beta 1.3.0 driver right now, since our setup requires authentication :-)
[17:17:08] <suri> i'm trying to use mongodb on django1.4.. as i know still there is no nonrel support for 1.4..
[17:20:08] <jrdn> weird question, should i cache things like "names" in memory when querying for an _id field? we have reports and are storing object ids (b/c names change often)… since i'm assuming it's in memory anyway, is it worth it to cache them anyway to get rid of what is like 40 queries on one report page?
[17:21:19] <jrdn> i.e.) http://docs.mongodb.org/manual/faq/fundamentals/#does-mongodb-handle-caching not sure if it's even worth it to build a cache layer right now
[17:24:56] <jrdn> and yeah, they are essentially a query on _id (indexed obviously) and returning just a name
[17:27:46] <_johnny> when querying on two fields in a collection, is it sufficient (albeit not optimal) to index only one of the fields?
[17:45:34] <_johnny> if you don't need documents, what is it you're using mongodb for? can you give an example?
[18:24:46] <Almindor> how would you go about importing a huge table (6 billion entries) from M$ SQL into MongoDB?
[18:24:59] <Almindor> we used freeTDS/odbc/python before but it's simply too slow for this huge thing
[18:26:20] <Almindor> I was wondering if there might be some non-obvious way of getting data into mongodb
[18:33:02] <Almindor> nodex: another issues it that it's not as simle as a push, I need to "re-link" IDs
[18:33:55] <NodeX> no way of replicating on the SQL machine while you import then dumping and moveing the data ?
[18:33:58] <Almindor> this is an events collection and they are connected to devices which in turn are connected to users :)
[18:39:46] <Almindor> I guess we'll try it with a month's set of data and see how long it takes on the mongo db machines to import it
[18:40:21] <NodeX> I know the longer a process runs the longer it takes and the more it slows down for some reason
[18:41:37] <NodeX> I remember somehting from a talk from the craigslist guy about it taking ages for him too
[18:48:18] <Almindor> this one? http://www.10gen.com/presentations/mongosf-2012/mongodb-at-craigslist-one-year-later
[19:22:08] <neil__g> hello. could anyone confirm for me that 2.0.7 is a drop-in replacement for 2.0.1? anything i need to be aware of in upgrading?
[19:23:40] <tsally> what are the performance implications of using safe writes. do drivers just poll getLastError() over and over again?
[19:36:55] <icedstitch> I mean, this is useful, http://media.mongodb.org/zips.json . i was wondering if there were other ones
[19:44:39] <kchodorow> icedstitch: here's an example using chess games http://www.kchodorow.com/blog/2012/01/26/hacking-chess-with-the-mongodb-pipeline/
[19:48:28] <estebistec> wow, so with pymongo .find({'a': re.compile('^AB')}) is much faster than .find({'a': {'$regex': ^AB'}})
[21:15:03] <tomlikestorock> why would my python script that connects to a replica set hang at the end of execution and not close the connection to mongo?
[21:18:11] <Bilge> How do you man the index of an embedded document unique only to the containing document and not the entire collection?
[21:22:36] <Bilge> In fact I don't even understand what it is that is unique about a compound index compising a root document's it and a child document's field
[21:23:43] <Bilge> I thought making a unique compound index compising the parent id and an embedded document's field would be a way to make that embedded document's field unique only in the context of the parent document but it does not
[21:40:49] <crudson1> Bilge: a unique index simply means that a combination of the keys specified must be unique in the collection
[21:41:41] <Bilge> So then I SHOULD NOT be able to SPECIFY that the parent ID and the embedded field is UNIQUE and yet STILL have the same embedded field with the same value in the same document!
[21:42:24] <Bilge> There is NOTHING UNIQUE about a compound index comprising a filed from the parent document and a field from an embedded document
[21:43:32] <Bilge> You can have a myriad of duplicate embedded field values regardless of the unique index constraint
[21:44:34] <Bilge> The only way to make it really unique is to create a single-field unique index for the field in the embedded document, e.g. { embedded.field : 1 } but this makes the field unique across all documents, not just the parent document to which the embedded document belongs
[21:45:38] <Bilge> Either this is a massive bug or I am massively misunderstanding a core concept of indexing
[21:47:35] <Bilge> ensureIndex({ _id: 1, 'embedded.field' : 1 }, { unique: true }) === ensureIndex({ _id: 1, 'embedded.field' : 1 }, { unique: false })
[21:49:10] <kida78> is there a bug in the mongodb shell that doesn't allow you to do queries while comparing to a float / double?
[22:21:44] <lacker> another question: in general, are there some types of error where a client like the ruby connection object will need to be recreated after that sort of error?
[22:32:43] <tomlikestorock> In my web app (pyramid based), is it good practice to open a replicaset connection on every request, and attach it to the request object, or should I open one for the app and attach a reference to every connection? I've also noticed that the replicasetconnection class causes my web app scripts to hang because it doesn't delete itself when the script is done running? How do I prevent this?
[22:33:38] <wereHamster> tomlikestorock: or pool connections and take one that is available or wait for one te become available
[22:34:07] <wereHamster> generally it's a bad idea to open a new connection to the database for each request.
[22:35:17] <tomlikestorock> wereHamster: right, that's what I figured. Okay. Now, what's the deal with the ReplicaSetConnection preventing my scripts from finishing?
[22:35:51] <wereHamster> no idea. Which programing languaeg you are using. Also, it's past midnight. I'm off.
[22:39:40] <Smith_> I am trying to run this example "https://github.com/mongodb/node-mongodb-native/blob/master/examples/simple.js" on Node.js v0.8.7 and getting this error "Error: TypeError: Cannot read property 'BSON' of undefined" on line 9's code. Any suggestions please?
[22:39:49] <winterpk> Can someone please tell me what the radius variable is supposed to be in this example: db.places.find({"loc" : {"$within" : {"$center" : [center, radius]}}})
[22:52:22] <Smith_> How is this related to Mongo? the variable units are defined by the data or definitions of the problem
[22:55:28] <Smith_> also check http://stackoverflow.com/questions/4143556/in-mongodbs-query-what-unit-is-center-radius-in
[23:31:44] <Smith_> What is the meaning of native parameter (true/false) when opening a DB? (or native in general in the context of mongodb?)
[23:49:20] <tomlikestorock> is there some way that my replicasetconnection from pymongo can close its own monitor thread upon deletion?