Menu
▾
▴
Re: [tauruslib-devel] client dependent memory leak
|
From: Zbigniew R. <zre...@ce...> - 2016-02-10 15:00:32
|
HI Roberto, Sorry for the late answer, but this week I could only dedicate some spare gaps of time to this problem. First of all, good that the re-installation of Tango made your memleak disappear. Then, what is strange to me are the results of my tests. I've done them at our beamline, which in theory is using the same version (latest 8.1.2 patched) in which you have seen the leak to disappear. But let me crosscheck it with my colleagues. Since I had no access to the beamline anymore, I had continued tests at my development environment. And I'm sure that I'm using the latest 8.1.2 patched version. I have not observer any memleak in TangoTest. I have also prepared a dummy DS (C++) with 10 attributes (5 scalar shorts and 5 scalar booleans) and I have checked them both on latest 8.1.2 patched version and on the 9.1.0. No leak was observed. I have it pending to check it with PyTango... But maybe as you said the latest 8.1.2 patched version solves it, and we do not use it in production... I will get back to you when I have more information. Cheers, Zibi On 02/10/2016 10:10 AM, Roberto Borghes wrote: > Hi Zibi, > the problem seems to be solved by installing the patched version of > Tango 8.1.2 available on http://www.tango-controls.org. > Yesterday I've compiled it and everything seems ok! > > Ciao, > Roberto > > > > On 02/02/2016 16:40, Zbigniew Reszela wrote: >> Hi Roberto, >> >> Thanks for this information. >> I will try to reproduce it with Tango 9 and TangoTest with the final >> objective to consult it on the Tango Forum. >> >> I will get back to you, >> Zibi >> >> >> On 02/02/2016 11:24 AM, Roberto Borghes wrote: >>> Hi Zibi, >>> in both cases I use taurusform from the command line with the device >>> name as argument, then I click the button "Show Device". >>> In this modality all the attributes are showed (also State and Status). >>> Currently we're not using events, so Taurus is polling the readings. >>> >>> Ciao, >>> Roberto >>> >>> >>> On 02/02/2016 10:42, Zbigniew Reszela wrote: >>>> Hi Roberto, >>>> >>>> So you start directly a taurusform with these 18 DevShort and 18 >>>> DevBoolean attributes, don't you? >>>> There is nothing else in the form? (no State, Status?). >>>> Do these attributes emit events? or it is Taurus who polls them? >>>> >>>> Good that you have reproduced it with TangoTest. Which attributes >>>> do you monitor in this case? All of the attributes of TangoTest? >>>> I will move to Tango 9 to check if the problem also exist there. >>>> >>>> Cheers, >>>> Zibi >>>> >>>> On 02/02/2016 10:00 AM, Roberto Borghes wrote: >>>>> Hi Zibi, >>>>> thank you for your analysis. I don't have any graph to show you >>>>> but I can confirm that the memory consumption is slow and constant. >>>>> Obviously the ratio depends on the number of attributes. The >>>>> device we're monitoring has 18 DevShort attributes and 18 >>>>> DevBoolean attributes. >>>>> Since yesterday, this device gained approximately 30 MB of memory >>>>> in 18 hours. >>>>> >>>>> But I've another information for you: the TangoTest device seems >>>>> to have the same problem. The memory consumption is slower but >>>>> does exists. >>>>> >>>>> Ciao, >>>>> Roberto >>>>> >>>>> >>>>> On 02/02/2016 09:26, Zbigniew Reszela wrote: >>>>>> Good morning! >>>>>> >>>>>> I have also observed slight increase in the memory - approx. 2 Mb >>>>>> per 14 hours. >>>>>> I explain you the setup which I used to monitor that. >>>>>> I simultaneously restarted two Python device servers (each of >>>>>> them exporting just one device), and their baseline memory >>>>>> footprint was: >>>>>> >>>>>> 24071 sicilia 20 0 178m 30m 16m S 1 1.2 0:03.13 python >>>>>> PyAlbaEm.py BL22_DI_EMET-01 >>>>>> 24373 sicilia 20 0 170m 30m 16m S 1 1.2 0:03.23 python >>>>>> PyAlbaEm.py BL22_DI_EMET-02 >>>>>> >>>>>> Then I have started a taurusform monitoring some models from the >>>>>> BL22_DI_EMET-01 instance. The other instance was not monitored by >>>>>> any GUI. >>>>>> These models were (in total 11): >>>>>> * the device itself >>>>>> * SCALAR DevDouble (for this one I have started polling every 3 >>>>>> seconds, with a very demanding absolute change threshold, making >>>>>> it emitting events almost every polling period) >>>>>> * 2 x SCALAR DevString attribute >>>>>> * SCALAR DevBoolean attribute >>>>>> * SCALAR DevLong attribute >>>>>> * 2 x SPECTRUM DevString attribute >>>>>> * SPECTRUM DevDouble attribute >>>>>> * State >>>>>> * Status >>>>>> >>>>>> After 14 hours the memory footprint was: >>>>>> >>>>>> 24071 sicilia 20 0 172m 32m 16m S 1 1.3 6:56.07 python >>>>>> PyAlbaEm.py BL22_DI_EMET-01 >>>>>> 24373 sicilia 20 0 170m 30m 16m S 1 1.2 6:18.53 python >>>>>> PyAlbaEm.py BL22_DI_EMET-02 >>>>>> >>>>>> See the attached plot with the memory consumption evolution. >>>>>> >>>>>> What is exactly your setup? Do you see similar leak? >>>>>> >>>>>> Thanks, >>>>>> Zibi >>>>>> >>>>>> >>>>>> >>>>>> On 02/01/2016 04:49 PM, Zbigniew Reszela wrote: >>>>>>> Hi Roberto! >>>>>>> >>>>>>> I'll move this discussion to the taurus-devel mailing list >>>>>>> (recently the Sardana and Taurus had separated their SourceForge >>>>>>> infrastructure) >>>>>>> >>>>>>> Here at Alba we use the following versions of the software: >>>>>>> >>>>>>> * server side: >>>>>>> ** OpenSUSE 11.3 32 bits >>>>>>> ** Tango 8.1.2.c + 4 patches from the Tango website >>>>>>> ** PyTango 8.1.4 >>>>>>> >>>>>>> * client side: >>>>>>> ** OpenSUSE 11.3 64 bits or OpenSUSE 12.1 64 bits >>>>>>> ** Tango 8.1.2.c + 4 patches from the Tango website >>>>>>> ** PyTango 8.1.4 >>>>>>> ** we try to follow the develop branch of Taurus (it is at least >>>>>>> 3.6.0) >>>>>>> >>>>>>> AFAIK we do not observe the memory leaks. But just in case I >>>>>>> will leave a taurusform with few attributes and devices running >>>>>>> overnight. >>>>>>> Tomorrow I will get back to you with the results. >>>>>>> >>>>>>> Meanwhile could you provide us more information: >>>>>>> * how many attributes / devices (models) do you have in your form. >>>>>>> * what type of attributes do you use? SPECTRUM, SCALAR, >>>>>>> DevFloat, DevLong, etc... >>>>>>> * do the attributes emit events? are they polled by the server? >>>>>>> * are your models coming from the default TANGO_HOST of the >>>>>>> client machine? or you specify the model names with the explicit >>>>>>> tango database and port? >>>>>>> * how do you choose them? specifying the models as the execution >>>>>>> parameters? or from the model chooser? or with the drag&drop >>>>>>> mechanism? or maybe you programmed taurusform GUI with the >>>>>>> hardcoded models? >>>>>>> >>>>>>> Cheers, >>>>>>> Zibi >>>>>>> >>>>>>> On 01/29/2016 05:49 PM, Roberto Borghes wrote: >>>>>>>> Hi everybody, >>>>>>>> here at Elettra we're facing a mysterious memory leak with some python >>>>>>>> device servers. >>>>>>>> The strangefact is that the memory leaks do appear only using taurus >>>>>>>> clients (like taurusform) >>>>>>>> , using atkpanel there is no memory increasing... >>>>>>>> >>>>>>>> If nobody faced this problem till now probably we've something wrong >>>>>>>> with our installation: >>>>>>>> >>>>>>>> - server side: Centos6 x64, Tango 8.1.2.c, PyTango 8.0.2 >>>>>>>> >>>>>>>> - client side: Centos7 x64, Tango 8.1.2.c, PyTango 8.1.6, taurus 3.6.0 >>>>>>>> >>>>>>>> Any information will be helpful. >>>>>>>> Thank you in advance, >>>>>>>> >>>>>>>> Roberto >>>>>>>> >>>>>>>> ------------------------------------------------------------------------------ >>>>>>>> Site24x7 APM Insight: Get Deep Visibility into Application Performance >>>>>>>> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month >>>>>>>> Monitor end-to-end web transactions and take corrective actions now >>>>>>>> Troubleshoot faster and improve end-user experience. Signup Now! >>>>>>>> http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140 >>>>>>>> _______________________________________________ >>>>>>>> Sardana-devel mailing list >>>>>>>> Sar...@li... >>>>>>>> https://lists.sourceforge.net/lists/listinfo/sardana-devel >>>>>>> >>>>>>> -- >>>>>>> >>>>>>> ALBA Synchrotron <http://www.albasynchrotron.es> >>>>>>> >>>>>>> Zbigniew Reszela >>>>>>> Controls Section - Computing Division >>>>>>> >>>>>>> ALBA SYNCHROTRON LIGHT SOURCE >>>>>>> Carrer de la Llum 2-26 | 08290 | Cerdanyola del Vallès| >>>>>>> Barcelona | Spain <http://www.albasynchrotron.es/AboutUs/Access> >>>>>>> (+34) 93 592 4407 >>>>>>> www.albasynchrotron.es| zre...@ce... >>>>>>> >>>>>>> **Please, do not print this e-mail unless it is absolutely >>>>>>> necessary. >>>>>>> **Si heu rebut aquest correu per error, us informo que pot >>>>>>> contenir informació confidencial i privada i que està prohibit >>>>>>> el seu ús. Us agrairíem que ho comuniqueu al remitent i >>>>>>> l'elimineu. Gràcies. >>>>>>> Si ha recibido este correo por error, le informo de que puede >>>>>>> contener información confidencial y privada y que está prohibido >>>>>>> su uso. Le agradeceré que lo comunique a su remitente y lo >>>>>>> elimine. Gracias. >>>>>>> If you have received this e-mail in error, please note that it >>>>>>> may contain confidential and private information, therefore, the >>>>>>> use of this information is strictly forbidden. Please inform the >>>>>>> sender of the error and delete the information received. Thank you. >>>>>>> >>>>>> >>>>>> -- >>>>>> >>>>>> ALBA Synchrotron <http://www.albasynchrotron.es> >>>>>> >>>>>> Zbigniew Reszela >>>>>> Controls Section - Computing Division >>>>>> >>>>>> ALBA SYNCHROTRON LIGHT SOURCE >>>>>> Carrer de la Llum 2-26 | 08290 | Cerdanyola del Vallès| Barcelona >>>>>> | Spain <http://www.albasynchrotron.es/AboutUs/Access> >>>>>> (+34) 93 592 4407 >>>>>> www.albasynchrotron.es| zre...@ce... >>>>>> >>>>>> **Please, do not print this e-mail unless it is absolutely necessary. >>>>>> **Si heu rebut aquest correu per error, us informo que pot >>>>>> contenir informació confidencial i privada i que està prohibit el >>>>>> seu ús. Us agrairíem que ho comuniqueu al remitent i l'elimineu. >>>>>> Gràcies. >>>>>> Si ha recibido este correo por error, le informo de que puede >>>>>> contener información confidencial y privada y que está prohibido >>>>>> su uso. Le agradeceré que lo comunique a su remitente y lo >>>>>> elimine. Gracias. >>>>>> If you have received this e-mail in error, please note that it >>>>>> may contain confidential and private information, therefore, the >>>>>> use of this information is strictly forbidden. Please inform the >>>>>> sender of the error and delete the information received. Thank you. >>>>>> >>>>> >>>> >>>> -- >>>> >>>> ALBA Synchrotron <http://www.albasynchrotron.es> >>>> >>>> Zbigniew Reszela >>>> Controls Section - Computing Division >>>> >>>> ALBA SYNCHROTRON LIGHT SOURCE >>>> Carrer de la Llum 2-26 | 08290 | Cerdanyola del Vallès| Barcelona | >>>> Spain <http://www.albasynchrotron.es/AboutUs/Access> >>>> (+34) 93 592 4407 >>>> www.albasynchrotron.es| zre...@ce... >>>> >>>> **Please, do not print this e-mail unless it is absolutely necessary. >>>> **Si heu rebut aquest correu per error, us informo que pot contenir >>>> informació confidencial i privada i que està prohibit el seu ús. Us >>>> agrairíem que ho comuniqueu al remitent i l'elimineu. Gràcies. >>>> Si ha recibido este correo por error, le informo de que puede >>>> contener información confidencial y privada y que está prohibido su >>>> uso. Le agradeceré que lo comunique a su remitente y lo elimine. >>>> Gracias. >>>> If you have received this e-mail in error, please note that it may >>>> contain confidential and private information, therefore, the use of >>>> this information is strictly forbidden. Please inform the sender of >>>> the error and delete the information received. Thank you. >>>> >>> >> >> -- >> >> ALBA Synchrotron <http://www.albasynchrotron.es> >> >> Zbigniew Reszela >> Controls Section - Computing Division >> >> ALBA SYNCHROTRON LIGHT SOURCE >> Carrer de la Llum 2-26 | 08290 | Cerdanyola del Vallès| Barcelona | >> Spain <http://www.albasynchrotron.es/AboutUs/Access> >> (+34) 93 592 4407 >> www.albasynchrotron.es| zre...@ce... >> >> **Please, do not print this e-mail unless it is absolutely necessary. >> **Si heu rebut aquest correu per error, us informo que pot contenir >> informació confidencial i privada i que està prohibit el seu ús. Us >> agrairíem que ho comuniqueu al remitent i l'elimineu. Gràcies. >> Si ha recibido este correo por error, le informo de que puede >> contener información confidencial y privada y que está prohibido su >> uso. Le agradeceré que lo comunique a su remitente y lo elimine. Gracias. >> If you have received this e-mail in error, please note that it may >> contain confidential and private information, therefore, the use of >> this information is strictly forbidden. Please inform the sender of >> the error and delete the information received. Thank you. >> > -- ALBA Synchrotron <http://www.albasynchrotron.es> Zbigniew Reszela Controls Section - Computing Division ALBA SYNCHROTRON LIGHT SOURCE Carrer de la Llum 2-26 | 08290 | Cerdanyola del Vallès| Barcelona | Spain <http://www.albasynchrotron.es/AboutUs/Access> (+34) 93 592 4407 www.albasynchrotron.es <http://www.albasynchrotron.es>| zre...@ce... <mailto:zre...@ce...> **Please, do not print this e-mail unless it is absolutely necessary. **Si heu rebut aquest correu per error, us informo que pot contenir informació confidencial i privada i que està prohibit el seu ús. Us agrairíem que ho comuniqueu al remitent i l'elimineu. Gràcies. Si ha recibido este correo por error, le informo de que puede contener información confidencial y privada y que está prohibido su uso. Le agradeceré que lo comunique a su remitente y lo elimine. Gracias. If you have received this e-mail in error, please note that it may contain confidential and private information, therefore, the use of this information is strictly forbidden. Please inform the sender of the error and delete the information received. Thank you. |
