During a recent automated update of our mongodb PRIMARY, when the PRIMARY stepped down it permanently went into a ROLLBACK state.
After several hours in the ROLLBACK state, there was still no rollback .bson file in the rollback directory in the mongodb database directory. That, and also this li...
If you open the OLAP server in SSMS you should be able to see them in the object explorer. You can configure the target DB on SSAS or enter it in the deployment wizard if you're deploying through that..
XMLA is a web service so you have to make up an XML file that has the web service request and the MDX query is a payload in the XML. You have to XML escape ;'s and so forth, but it can be quite useful for various things.
You can directly query the database. In theory you could make up a series of XMLA requests to construct a cube, but that's far too much like hard work.
Unfortunately MS never really refined DDL on SSAS to the point where you could easily knock up a file that built a cube in a text editor and run it in the way you can with SQL. Technically possible but far more trouble than it's worth.
@CadeRoux I'm a big fan of doing precisely that - all access to the data goes through a single well-defined transactional API. It allows you to (loosely) enforce all sorts of constraints you can't do with DRI and your API transactions map to business transations/rules naturally
@AaronBertrand I know it wasn't the point, but these 'gimme teh code' questions don't fly on SO, unless it's a really good question. More often than not, if you think your question is that good, that's your first mistake. By default, it's not.
@Phil Be the diamond on the other side handling flags knowing that what you're getting is a small fraction of the overall crap on the site (even when the queue is cleared, currently we have 380+ flags). =)
@JackDouglas It seems like an OO programmer who strives for decoupling and ability to refactor would AT LEAST want to use views so they COULD refactor if they wanted to.
@JackDouglas It seems disingenuous to me to talk about using ORMs and hibernate for the ability to change database platform agnostically and then not accept that after the level you hand off to the database there are also some kind of developer entities similar to yourself who might want to refactor and perhaps there should be a well-defined interface at that part of the system instead of "ALL TABLES ALL THE TIME".
@JNK - see where it says database in the properties dialog, enter another database name and deploy. It will create a new DB on the SSAS server. Then you can set up the DB connection details to point to the other SQL Server source DB.
@JackDouglas Well, although these past two days, extremely busy. I'm the de-facto DBA at my client, and well, that's not a good thing, haha. Given I have a task list as long as both my arms, being DBA for their data warehouse is an annoying side job which I don't get paid for. But really, I can't complain. I'm happy, wife is happy. Trying to have kids now. Exciting times.
@CadeRoux I use views for queries and functions for transactions - refactoring is a bonus, mostly I just want control. OO principles generally translate quite poorly to the database layer imo - or in some cases not poorly but just in very different ways that are almost unrecognisable to an OO programmer at first
@casperOne well I'm glad if you are able to get help here every now and again :-) I guess we should expect to see a lot less of you soon if your plans bear fruit...
@JNK That's SSMS opened on a SSAS DB. If you open the 'data sources node' and right click on the data source. Select properties and it brings up that dialog. You can change the connection string there.
@JackDouglas Nah, don't plan on it. I'll still have to work, and I get most of my SO/SE modding done at work =) I'll be around. Can't forget the peeps. =)
Deploy it to a new DB from BIDS, then open the new DB in SSMS and change the connection string for the data source to point to your other customer's database. Then do a full process on the cubes.
End result: two identical SSAS cubes pointed at different sets of source data.
@JackDouglas My point is that there is no difference in perspective between OO programmers and DB programmers at the level of engineering principles. We all want decoupling, we all want well-defined interfaces, we all want layers and reliable independence of refactoring within layers. So why does the "dumb datastore" approach or the "everything in code and we'll just run a lot of loops" still persist?
Software isn't designed these days. It's rapidly prototyped ("agile") in iterations based on (often loose) requirements. There's no time to design a datastore and then build on top of it. Much easier to write some shit Java and have it manage the DDL & add random columns whenever the object models change
@Phil I develop like that too to a degree (it's hard doing it any other way when you work for SMBs), but still build the datastore first and code on top. I believe it is quicker.
@JackDouglas I dispute that - they'll jump all over node.js or Ruby or jQuery or whatever the latest flavor is - they love to jump on the next tool bandwagon. It's not because SQL is declarative - they jump all over declarative things like XML and CSS. It's not the strictness - they'll love to limit themselves with frameworks like Rails.
@swasheck Is SQL just too old? I know it's inconsistent between platforms. I know it's a lot different from most languages - but it really is NO HARDER than modern JavaScript usage, which has jumped right into the functional deep end compared to when it first was used in the mid-90s.
@Phil A LOT of stuff is moving to JavaScript. It should be in databases any day now - just like CLR SPs or Java stuff in Oracle. Does SSIS support it in scripts yet or just VB still?
@ConcernedOfTunbridgeWells Last version I used SSIS script tasks (2005?), it only had VB. With JSON and the proliferation of various js libraries, I think it's only a matter of time before JS engines get into databases. Too bad the I18n support in js (and its builtin date) is so poor. It has some nice concepts to go with all its annoying quirks which will probably never get eradicated.
@CadeRoux Script tasks are .Net based. There is an activeX scripting component that uses WSH - a hangover from older pre-2005 DTS - but it's a different type of component.
You could probably bolt whatever the open-source mozilla one (spidermonkey?) into PostgreSQL without too much fuss - it's already got embedded perl, python, tcl and pl/pgsql interpreters.
SQL Server has a .Net CLR runtime built in, so you can load pretty much anything built in a language that supports .Net.
@JackDouglas JS never really got much traction in OSS circles outside of web browsers. I think it's up against a fair bit of incumbency in Python and Perl.
@ConcernedOfTunbridgeWells indeed. if you look at @NickChammas youtube link to the mongodb video (which is a hilarious view), there's a related video that talks about how only an idiot would use node.js. he got flamed in the comments
Douglas Schmid's web page (if it's still up) has a whole pile of stuff on this type of architecture - they did a C++ framework called ACE back in the 1990s.
@swasheck I like it. @Phil's just started tinkering with it and seems to be quite chuffed with his results as well.
A very old-school C++ server framework, originally designed for telecoms applications. However, if you want efficient and fast, ACE certainly qualifies for that.
I think there's an async server framework in BOOST as well now - maybe called BOOST:Async or something like that.
Given the amount of CPU you can actually put on a modern server, I wonder how much you could scale up something like ACE or BOOST:Async without having to resort to a distributed architecture.
Although the majority of my work is in SQL and C#, I've recently become very interested in functional programming and have become more interested in the modern idiomatic JavaScript (as opposed to what I learned and used for a lot of HTML DOM manipulation in the late 90s). I find that things like Haskell are much closer to SQL with the declarative and lazy evaluation where you effectively declare and see what the runtime does as opposed to the imperative model of other languages.
@CadeRoux Maybe you should try tinkering with erlang at some point.
Although most of my memories of functional programming stem from when I worked as a tutor on a first year functional programming course. Trying to make sense of lab work done by students who really, really just didn't get functional programming at all was quite a painful exercise.
Not only did you have to think of the problem in functional terms, you got to try and deconstruct the misconceptions of someone who just hadn't twigged onto it at all ans was completely barking up the wrong tree.
This was the compulsory course to get into stage 2. It was a most effective filter to keep people out of the CS department who really had no business being there.
"Get your head around functional programming and we'll let you into Stage 2"
@CadeRoux On a side note, is MS still punting F#? It's a descendent of the ML family and plays nicely with other .Net languages.
@ConcernedOfTunbridgeWells The whole no-side-effects thing is a very interesting constraint which has a lot of payoffs. Yes, I think I'm going to invest some time looking at F#, since it obviously fits well in the .NET ecosystem. Also, Mono since I'm interested in Android and haven't really done any mobile native stuff since the Psion days.
@MarkStoreySmith @NickChammas @swasheck Thanks for helping me earlier. I've posted the following question as a result (and I'm sure one or any of the people on DBA can answer it way better than I can). Thanks again for the pointers and help.
I have a large (in the tens of millions of records) database that I am going to perform a full database backup on.
However, the database is large enough that transactions can start before and during, as well as commit during and after the backup takes place.
For example:
T0 = Transaction A sta...
During a recent automated update of our mongodb PRIMARY, when the PRIMARY stepped down it permanently went into a ROLLBACK state.
After several hours in the ROLLBACK state, there was still no rollback .bson file in the rollback directory in the mongodb database directory. That, and also this li...
I have a large (in the tens of millions of records) database that I am going to perform a full database backup on.
However, the database is large enough that transactions can start before and during, as well as commit during and after the backup takes place.
For example:
T0 = Transaction A sta...
