July 2013 - Pharo-users - lists.pharo.org

Re: [Pharo-users] MongoDB open close in production (singleton or not)

by Norbert Hartl

Am 18.07.2013 um 15:06 schrieb Sabine KnÃ¶fel <sabine.knoefel(a)gmail.com>: > Hi Norbert and Mariano, > > thanks for the important input. > I will use voyage now, just loading :-) > > BTW: it is really great, getting answers in the forum so quick and > profound!!! I am very happy about that. > In my case it is pure selfishness! I think answering a question in a well-founded way is a great of learning stuff. So I'm not sure you gains more ;) Anyway, you're welcome! Norbert > Sabine > > On Thu, Jul 18, 2013 at 2:57 PM, Norbert Hartl [via Smalltalk] > <[hidden email]> wrote: > > > > > Am 18.07.2013 um 12:33 schrieb Sabine KnÃ¶fel <[hidden email]>: > > > >> Hi, > >> > >> Currently, I create one single Instance of Mongo when starting the Image. > >> > >> | theRoot | > >> theRoot := Mongo default. > >> theRoot open. > >> > >> and use it for all requests. In my development environment that works > >> fine. > >> > >> My question is: is this the right way and will it work for production? > > > > No, certainly not. If you restart mongo db e.g. due a system update the > > connection from the image would be stale serving you errors. > >> Or is it better to create one Mongo instance for each request (and close > >> it > >> after the request)? > > That is theoretically the safest thing but very expensive. Opening a > > connection always takes time and consume quite some resources. So reuse is > > king. > >> And, btw: what would happen if I would open many Mongo connections but > >> never > >> close any? > > You would run out of external semaphores in the image. The image is > > "somewhat broken" in this regard because it cannot expand the semaphore > > table. As each connection uses three semaphores you are running out of > > semaphores after 85 requests. > > > >> Or better one Instance per user session? > >> > > That can be a good idea but it doesn't solve your actual problem. > > > > You also need to take care of concurrent requests. The line protocol to the > > mongo database needs to be aligned and it is binary. If more threads try to > > write on the same connection the connection will break. I'm not sure if this > > is still possible in the newest MongoTalk driver nor do I know if I managed > > it to release my thread safe mongo instance. > > > > The pool Mariano is talking about sounds like a good idea. Otherwise wrap > > your store method with an exception handler to reopen a connection on > > failure. > > > > Norbert > > > > > > > > > > ________________________________ > > If you reply to this email, your message will be added to the discussion > > below: > > http://forum.world.st/MongoDB-open-close-in-production-singleton-or-not-tp4… > > To unsubscribe from MongoDB open close in production (singleton or not), > > click here. > > NAML > > View this message in context: Re: MongoDB open close in production (singleton or not) > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.

July 18, 2013

Re: [Pharo-users] MongoDB open close in production (singleton or not)

by Sabine KnÃ¶fel

Hi Norbert and Mariano, thanks for the important input. I will use voyage now, just loading :-) BTW: it is really great, getting answers in the forum so quick and profound!!! I am very happy about that. Sabine On Thu, Jul 18, 2013 at 2:57 PM, Norbert Hartl [via Smalltalk] <ml-node+s1294792n4699354h69(a)n4.nabble.com> wrote: > > Am 18.07.2013 um 12:33 schrieb Sabine KnÃ¶fel <[hidden email]>: > >> Hi, >> >> Currently, I create one single Instance of Mongo when starting the Image. >> >> | theRoot | >> theRoot := Mongo default. >> theRoot open. >> >> and use it for all requests. In my development environment that works >> fine. >> >> My question is: is this the right way and will it work for production? > > No, certainly not. If you restart mongo db e.g. due a system update the > connection from the image would be stale serving you errors. >> Or is it better to create one Mongo instance for each request (and close >> it >> after the request)? > That is theoretically the safest thing but very expensive. Opening a > connection always takes time and consume quite some resources. So reuse is > king. >> And, btw: what would happen if I would open many Mongo connections but >> never >> close any? > You would run out of external semaphores in the image. The image is > "somewhat broken" in this regard because it cannot expand the semaphore > table. As each connection uses three semaphores you are running out of > semaphores after 85 requests. > >> Or better one Instance per user session? >> > That can be a good idea but it doesn't solve your actual problem. > > You also need to take care of concurrent requests. The line protocol to the > mongo database needs to be aligned and it is binary. If more threads try to > write on the same connection the connection will break. I'm not sure if this > is still possible in the newest MongoTalk driver nor do I know if I managed > it to release my thread safe mongo instance. > > The pool Mariano is talking about sounds like a good idea. Otherwise wrap > your store method with an exception handler to reopen a connection on > failure. > > Norbert > > > > > ________________________________ > If you reply to this email, your message will be added to the discussion > below: > http://forum.world.st/MongoDB-open-close-in-production-singleton-or-not-tp4… > To unsubscribe from MongoDB open close in production (singleton or not), > click here. > NAML -- View this message in context: http://forum.world.st/MongoDB-open-close-in-production-singleton-or-not-tp4… Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.

July 18, 2013

Re: [Pharo-users] MongoDB open close in production (singleton or not)

by Norbert Hartl

Am 18.07.2013 um 12:33 schrieb Sabine KnÃ¶fel <sabine.knoefel(a)gmail.com>: > Hi, > > Currently, I create one single Instance of Mongo when starting the Image. > > | theRoot | > theRoot := Mongo default. > theRoot open. > > and use it for all requests. In my development environment that works fine. > > My question is: is this the right way and will it work for production? No, certainly not. If you restart mongo db e.g. due a system update the connection from the image would be stale serving you errors. > Or is it better to create one Mongo instance for each request (and close it > after the request)? That is theoretically the safest thing but very expensive. Opening a connection always takes time and consume quite some resources. So reuse is king. > And, btw: what would happen if I would open many Mongo connections but never > close any? You would run out of external semaphores in the image. The image is "somewhat broken" in this regard because it cannot expand the semaphore table. As each connection uses three semaphores you are running out of semaphores after 85 requests. > Or better one Instance per user session? > That can be a good idea but it doesn't solve your actual problem. You also need to take care of concurrent requests. The line protocol to the mongo database needs to be aligned and it is binary. If more threads try to write on the same connection the connection will break. I'm not sure if this is still possible in the newest MongoTalk driver nor do I know if I managed it to release my thread safe mongo instance. The pool Mariano is talking about sounds like a good idea. Otherwise wrap your store method with an exception handler to reopen a connection on failure. Norbert

July 18, 2013

Re: [Pharo-users] MongoDB open close in production (singleton or not)

by Mariano Martinez Peck

Hi Sabine, take a look to Voyage, and its class VOMongoSessionPool. I think a nice way of dealing with what you ask is using the pool and each time you need to do something with the DB, you do it inside a block to #withDatabase: myPool withDatabase: [:database| self doSomething ] you also have #withSession instead of #withDatabase. btw, I guess we can move the pool class from voyage to mongo since it's decoupled. Cheers, On Thu, Jul 18, 2013 at 7:33 AM, Sabine KnÃ¶fel <sabine.knoefel(a)gmail.com>wrote: > Hi, > > Currently, I create one single Instance of Mongo when starting the Image. > > | theRoot | > theRoot := Mongo default. > theRoot open. > > and use it for all requests. In my development environment that works fine. > > My question is: is this the right way and will it work for production? > Or is it better to create one Mongo instance for each request (and close it > after the request)? > And, btw: what would happen if I would open many Mongo connections but > never > close any? > Or better one Instance per user session? > > Sabine > > (I posted this in Frameworks forum, possibly better here) > > > > -- > View this message in context: > http://forum.world.st/MongoDB-open-close-in-production-singleton-or-not-tp4… > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. > > -- Mariano http://marianopeck.wordpress.com

July 18, 2013

New post on Voyage: Advanced queries

by Esteban Lorenzano

Hi, I just finished a section for the PharoForTheEnterprise book and I put it also as a blog post: http://smallworks.eu/web/blog/2013-07-18-Voyage-advanced-queries cheers, Esteban

July 18, 2013

MongoDB open close in production (singleton or not)

by Sabine KnÃ¶fel

Hi, Currently, I create one single Instance of Mongo when starting the Image. | theRoot | theRoot := Mongo default. theRoot open. and use it for all requests. In my development environment that works fine. My question is: is this the right way and will it work for production? Or is it better to create one Mongo instance for each request (and close it after the request)? And, btw: what would happen if I would open many Mongo connections but never close any? Or better one Instance per user session? Sabine (I posted this in Frameworks forum, possibly better here) -- View this message in context: http://forum.world.st/MongoDB-open-close-in-production-singleton-or-not-tp4… Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.

July 18, 2013

Re: [Pharo-users] SerialPlugin

by Bernat Romagosa

Still trying without success... Does anyone have an image, a VM and a SerialPlugin working in their machine that they can share with me? At this point I don't care too much on which OS, I'll find a machine here at work. The thing is I wanted to start developing right away, and this little obstacle is holding me back... Thanks a lot! Bernat. 2013/7/17 Bernat Romagosa <tibabenfortlapalanca(a)gmail.com> > Yep, I found that thread and tried all possible combinations: > > '/dev/ttyUSB0' > '/dev//ttyUSB0' > '/dev/ttyusb0' > '/dev//ttyusb0' > 'ttyUSB0' > 'ttyusb0' > > and none worked... :( > > > 2013/7/17 Esteban Lorenzano <estebanlm(a)gmail.com> > >> It is indeed annoying and obviously buggy. >> I kinda "fixed" it last year but since I didn't had any serial port stuff >> to test it, it was just in abstract (Luc was using it, thought) >> >> There was a report a couple of weeks ago to try using double slash. >> >> SerialPort new openPort: '/dev//ttyUSB0' >> >> >> >> Esteban >> >> On Jul 17, 2013, at 1:47 PM, Bernat Romagosa < >> tibabenfortlapalanca(a)gmail.com> wrote: >> >> So I compiled two VMs, one with the SerialPlugin as an internal plugin, >> the other one as an external plugin. >> >> On a Debian box, I connect an Arduino Diecimila board and I get its >> handler at /dev/ttyUSB0. However, on both VMs with fresh Pharo 2.0 images: >> >> SerialPort new openPort: '/dev/ttyUSB0' >> >> >> Fails with a generic error. >> >> What's wrong with SerialPort? The "cannot open port" error message is >> about as informative as "sorry, not gonna happen" :( >> >> >> 2013/7/17 Bernat Romagosa <tibabenfortlapalanca(a)gmail.com> >> >>> Great! I was missing the gl and alsa headers. All done now, thanks a lot >>> :) >>> >>> >>> 2013/7/17 Guillermo Polito <guillermopolito(a)gmail.com> >>> >>>> >>>> On Wed, Jul 17, 2013 at 12:39 PM, Bernat Romagosa < >>>> tibabenfortlapalanca(a)gmail.com> wrote: >>>> >>>>> Okay, it seems the patch is already included in the latest VMMaker >>>>> image, so no need to apply it. >>>>> >>>> >>>> That's nice :). Anyone with grants in the issue tracker can close it?? >>>> :D >>>> >>>> >>>>> >>>>> Still, having executed the script, I get yet another error, this time >>>>> related to gl.h (why does it need it? I left B3DAcceleratorPlugin out): >>>>> >>>>> [ 95%] Building C object >>>>> CMakeFiles/CogVM.dir/home/bromagosa/blessed/platforms/unix/vm/sqUnixMain.c.o >>>>> In file included from >>>>> /home/bromagosa/blessed/platforms/unix/vm/SqDisplay.h:16:0, >>>>> from >>>>> /home/bromagosa/blessed/platforms/unix/vm/sqUnixMain.c:140: >>>>> /home/bromagosa/blessed/platforms/unix/plugins/B3DAcceleratorPlugin/sqUnixOpenGL.h:21:22: >>>>> fatal error: GL/gl.h: El fitxer o directori no existeix >>>>> compilation terminated. >>>>> make[2]: *** >>>>> [CMakeFiles/CogVM.dir/home/bromagosa/blessed/platforms/unix/vm/sqUnixMain.c.o] >>>>> Error 1 >>>>> make[1]: *** [CMakeFiles/CogVM.dir/all] Error 2 >>>>> make: *** [all] Error 2 >>>>> >>>>> >>>> Hmm, which configuration are you using to copile? I think there is a >>>> UnixNoGLConfiguration or something like that... So either you install the >>>> gl libraries or use the no gl configuration to generate the sources :). >>>> In any case, I really don't know why there is that dependency with >>>> opengl... >>>> >>>> >>>>> >>>>> 2013/7/17 Bernat Romagosa <tibabenfortlapalanca(a)gmail.com> >>>>> >>>>>> ouch! I forgot to execute the script this time, and didn't install >>>>>> the patch either.. trying again, thanks! :) >>>>>> >>>>>> >>>>>> 2013/7/17 Guillermo Polito <guillermopolito(a)gmail.com> >>>>>> >>>>>>> Hi! >>>>>>> >>>>>>> before compiling you have to, inside the build folder, execute >>>>>>> codegen-scripts/extract-commit-info.sh >>>>>>> >>>>>>> Besides, I was looking at luc's tutorial and it lacks one step to >>>>>>> have serial plugin built: install the patch he made [1]. Did you do that? I >>>>>>> added a comment in the blog entry just in case... >>>>>>> >>>>>>> Guille >>>>>>> >>>>>>> [1] https://code.google.com/p/cog/issues/detail?id=119&q=luc >>>>>>> >>>>>>> >>>>>>> On Wed, Jul 17, 2013 at 12:18 PM, Bernat Romagosa < >>>>>>> tibabenfortlapalanca(a)gmail.com> wrote: >>>>>>> >>>>>>>> I found out how to leave B3DAcceleratorPlugin out, but now I get >>>>>>>> this error almost at the end of the compilation process: >>>>>>>> >>>>>>>> [ 95%] Building C object >>>>>>>> CMakeFiles/CogVM.dir/home/bromagosa/blessed/platforms/unix/vm/sqUnixMain.c.o >>>>>>>> In file included from >>>>>>>> /home/bromagosa/blessed/platforms/unix/vm/sqUnixMain.c:41:0: >>>>>>>> /home/bromagosa/blessed/platforms/Cross/vm/sqSCCSVersion.h:17:54: >>>>>>>> fatal error: vmVersionInfo.h: El fitxer o directori no existeix >>>>>>>> compilation terminated. >>>>>>>> make[2]: *** >>>>>>>> [CMakeFiles/CogVM.dir/home/bromagosa/blessed/platforms/unix/vm/sqUnixMain.c.o] >>>>>>>> Error 1 >>>>>>>> make[1]: *** [CMakeFiles/CogVM.dir/all] Error 2 >>>>>>>> make: *** [all] Error 2 >>>>>>>> >>>>>>>> >>>>>>>> "El fitxer o directori no existeix" is Catalan for "The file or >>>>>>>> directory does not exist". >>>>>>>> >>>>>>>> How comes this header isn't there? >>>>>>>> >>>>>>>> >>>>>>>> 2013/7/17 Bernat Romagosa <tibabenfortlapalanca(a)gmail.com> >>>>>>>> >>>>>>>>> Ok, it seems the problem was not having the sources file in the >>>>>>>>> same directory as the image. >>>>>>>>> >>>>>>>>> Anyway, now I've another problem when trying to compile. cmake >>>>>>>>> runs fine, but make fails with the following error: >>>>>>>>> >>>>>>>>> Scanning dependencies of target B3DAcceleratorPlugin >>>>>>>>> [ 57%] Building C object >>>>>>>>> B3DAcceleratorPlugin/CMakeFiles/B3DAcceleratorPlugin.dir/home/bromagosa/blessed/src/plugins/B3DAcceleratorPlugin/B3DAcceleratorPlugin.c.o >>>>>>>>> In file included from >>>>>>>>> /home/bromagosa/blessed/src/plugins/B3DAcceleratorPlugin/B3DAcceleratorPlugin.c:36:0: >>>>>>>>> /home/bromagosa/blessed/platforms/Cross/plugins/B3DAcceleratorPlugin/B3DAcceleratorPlugin.h:8:19: >>>>>>>>> fatal error: GL/gl.h: El fitxer o directori no existeix >>>>>>>>> compilation terminated. >>>>>>>>> make[2]: *** >>>>>>>>> [B3DAcceleratorPlugin/CMakeFiles/B3DAcceleratorPlugin.dir/home/bromagosa/blessed/src/plugins/B3DAcceleratorPlugin/B3DAcceleratorPlugin.c.o] >>>>>>>>> Error 1 >>>>>>>>> make[1]: *** >>>>>>>>> [B3DAcceleratorPlugin/CMakeFiles/B3DAcceleratorPlugin.dir/all] Error 2 >>>>>>>>> make: *** [all] Error 2 >>>>>>>>> >>>>>>>>> >>>>>>>>> Any ideas? >>>>>>>>> >>>>>>>>> (For the record, I do not need the B3DAcceleratorPlugin, so I >>>>>>>>> guess it could just be left out, but I don't know how does one do that >>>>>>>>> either) >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> 2013/7/16 Friedrich Dominicus <frido(a)q-software-solutions.de> >>>>>>>>> >>>>>>>>>> Bernat Romagosa <tibabenfortlapalanca(a)gmail.com> writes: >>>>>>>>>> >>>>>>>>>> > Hi Friedrich, >>>>>>>>>> > >>>>>>>>>> > My machine is already a 32 bits one... here's the output of ldd: >>>>>>>>>> > >>>>>>>>>> > linux-gate.so.1 => (0xb7782000) >>>>>>>>>> > libm.so.6 => /lib/i386-linux-gnu/i686/cmov/libm.so.6 >>>>>>>>>> > (0xb7746000) >>>>>>>>>> > libdl.so.2 => /lib/i386-linux-gnu/i686/cmov/libdl.so.2 >>>>>>>>>> > (0xb7742000) >>>>>>>>>> > libpthread.so.0 => >>>>>>>>>> > /lib/i386-linux-gnu/i686/cmov/libpthread.so.0 (0xb7728000) >>>>>>>>>> > libc.so.6 => /lib/i386-linux-gnu/i686/cmov/libc.so.6 >>>>>>>>>> > (0xb75c5000) >>>>>>>>>> > /lib/ld-linux.so.2 (0xb7783000) >>>>>>>>>> > >>>>>>>>>> > Thanks! :) >>>>>>>>>> Well that looks fine to me. Sorry that I could not help >>>>>>>>>> >>>>>>>>>> Regards >>>>>>>>>> Friedrich >>>>>>>>>> >>>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Bernat Romagosa. >>>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> -- >>>>>>>> Bernat Romagosa. >>>>>>>> >>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> Bernat Romagosa. >>>>>> >>>>> >>>>> >>>>> >>>>> -- >>>>> Bernat Romagosa. >>>>> >>>> >>>> >>> >>> >>> -- >>> Bernat Romagosa. >>> >> >> >> >> -- >> Bernat Romagosa. >> >> >> > > > -- > Bernat Romagosa. > -- Bernat Romagosa.

July 18, 2013

Re: [Pharo-users] installing Seaside on Pharo 2.0

by StÃ©phane Ducasse

can you retry because it works from here. On Jul 18, 2013, at 7:13 AM, sminni <sm(a)planage.com> wrote: > Thanks ... got the Zinc server to work > > but could not get to smalltalkhub.com > I just get a "Loading please wait" message that sits till I give up > What is the way to get the HOSeaside Package > > regards > Sanjay > > > > ----- > --- > Regards, Sanjay > -- > View this message in context: http://forum.world.st/installing-Seaside-on-Pharo-2-0-tp4698685p4699247.html > Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com. >

July 18, 2013

Re: [Pharo-users] development image memory use: 180MB in Pharo 2 vs. 40MB in Pharo 1.4.....

by StÃ©phane Ducasse

thanks paul for the information. We should pay attention to this. May be we will have to have one release of Pharo focusing on not improving the libraries but its runtime behavior. The problem is that we need enough room to move and improve the infrastructure. > I've been developing in the same Pharo2 image for a couple weeks and its > now 180MB. I use the same development habits as in Pharo 1.4 but there > the image would only be about 40MB. > > Using SpaceTally I can see that I have 40 MB of Arrays. And 10MB of > Bitmaps. 8MB of Floats, Strings, & OrderedCollections. Strange. We should check may be some memory leaks. > > My domain objects comprise about 20MB of the image. > > > Is there a way to check for memory leaks over a long period of time? > > Thanks > > Paul >

July 18, 2013

Re: [Pharo-users] development image memory use: 180MB in Pharo 2 vs. 40MB in Pharo 1.4.....

by StÃ©phane Ducasse

On Jul 18, 2013, at 8:04 AM, Marcus Denker <marcus.denker(a)inria.fr> wrote: > > On Jul 17, 2013, at 9:59 PM, Paul DeBruicker <pdebruic(a)gmail.com> wrote: > >> I've been developing in the same Pharo2 image for a couple weeks and its >> now 180MB. I use the same development habits as in Pharo 1.4 but there >> the image would only be about 40MB. >> > > I already saw that the current 3.0 download has 110.000 instances of arrayâ¦ > I wondered if that makes sense. We should have a look. > >> Using SpaceTally I can see that I have 40 MB of Arrays. And 10MB of >> Bitmaps. 8MB of Floats, Strings, & OrderedCollections. >> >> My domain objects comprise about 20MB of the image. >> > > Can you run > > ImageCleaner cleanUpForRelease > > this calls e.g. all the #cleanUp methods of all classes. > > > (We should think about how to call this (or a subset) more regularly > in normal operation) > > Marcus > > > >

July 18, 2013