I'm happy to let the Java PP testing slide to 1.5.0 There are some recent improvements in the ruby PP that I need to implement. * sakaidocs - (easy, call out to wkhtmltopdf) * image previews in the same format as the original
Erik On Tue, Jul 24, 2012 at 10:18 AM, Kent Fitzgerald <kentf...@umich.edu> wrote: > Several questions/comments. > There has already been 1.4.1. release proposed for immediately following > 1.4.0 that would be isolated to code reformatting . Which would take > precedence? > > We should definitely do a bug bash. One of the dangers of doing a bug bash > focused on the preview processor is that we'll likely have people uploading > hundreds of files each. Subjectively, this could give the impression of > decreased performance just because we're hitting it much harder. > > More importantly, in addition to the bug bash, we need to do controlled > tests on processing time on different data types. I'd like to break it down > by file types and have truly controlled tests, in addition to different file > types we'll need files of varying sizes to compare performance not just on > quantity but on complexity. This needs to be compared to the performance of > the current implementation. > > I think we all agree that this is an important feature that we shouldn't try > to rush out the door. > > I have to read back through the thread, but is there set-up documentation? > Currently we have a section on the OAE Configuration and Deployment page [1] > for the preview processor. It's contains multiple supporting external links > that have proven confusing for many people trying to get preview processor > running locally. We'll need to make sure we have adequate documentation. > > As a side note, I will be out of the office starting this Friday through > next week. > > > [1] > https://confluence.sakaiproject.org/display/3AK/OAE+Configuration+and+Deployment > > > > -- > Kent Fitzgerald > > On Tuesday, July 24, 2012 at 9:51 AM, Nicolaas Matthijs wrote: > > Looks like this has been hanging around on list for a while now, and we > should probably try to move it forwards. > > The maintainability criterion can only be determined by a code review, which > is standard practice. However, as this is proving to be such a critical > feature in production, I'd suggest that we do a separate bugbash to evaluate > its performance, ease of setup (running from a separate machine) and most > importantly functional equivalence. > > When doing this, Kent can give his assessment of the ease of setup and the > bugbashers can determine functional equivalence. We should also try to have > it re-process the dummy content we usually bugbash with. > > If this all sounds good, I'd like to go ahead with this as soon as possible > and run a bugbash straight after the 1.4.0 release with all of this set up. > If the implementation survives the bugbash, it can be reviewed and merged. > > Does that sound reasonable? > > Thanks, > Nicolaas > > > > On 23 Jul 2012, at 07:42, Carl Hall wrote: > > Lance, I think the work is already split the way you suggest given what I > know about what Erik has done (rewrite in Java) and what's left (add JMS). > Adding message queue capabilities should not hold back reviewing the > proposed changes. > > I would say that it needs to meet these opening criteria for my general > acceptance: > > * Be functionally equal with the current solution > * A combination of performance and maintainability > * Perform can be no worse overall. There might be different hotspots in > the java version than the current ruby solution but there shouldn't be > anything exponentially worse. Overall, the java version has to perform at > least as good and hopefully better. Memory usage, overall processing time, > resource usage (iops, disc reads, caching) should all be considered. > * Be more maintainable than the Ruby solution. The current code has had > very little cleaning and is not very readable. This includes using > externally available libraries where possible. We shouldn't be maintaining > plumbing not inherent to our domain. > * Easier to setup. Though our current setup for the ruby PP is known to be > problematic, we at least are accustomed to it. The proposed solution has got > to be more straightforward and less fragile. > > The numbers I've seen from some preliminary testing showed the Java impl to > take exponentially *less* time to process pdfs and was faster than the ruby > PP in every test. It's an OSGi bundle and written in Java like the rest of > our project which makes it easier to setup and maintain as we write far more > java code than ruby. I believe there's also already a setup available to run > the java PP as a standalone server. > The Java version introduces a topia term extractor bundle which is a port > from the Python version. This is a point of maintenance to consider but the > python code has changed in years. It's a common impl for other languages to > port but there wasn't a java version around. I would like to see this code > find a permanent home in a relative OSS project. At the very least it should > be maintained apart from OAE core to make it available to a broader > audience. > > +1 to getting this code wrapped up and reviewed. > > On Wed, Jul 18, 2012 at 1:51 PM, Christian Vuerings > <vueringschrist...@gmail.com> wrote: > > I'm not sure whether this is already part of the criteria list or not, but > what about CPU/Memory usage? > Is there a way we can measure that and compare it to the current ruby based > PP? > When I currently run the ruby PP locally, it's usually one of the processes > that uses the most resources. > > One other thing I'm curious about is how well it will compress/handle the > different file formats (png/jpg/gif/psd) > > These are just 2 things that I'm interested in since they (can) have an > impact on the overall performance. > > > - Christian > > On Jul 18, 2012, at 12:41 PM, Lance Speelmon wrote: > > Does anyone have an opinion about adopting the new java based PP? > Specifically can you articulate acceptance criteria for such an adoption? > e.g. > > Must support same preview behaviors as existing ruby-based PP. > Must pass QA with all blocker and critical items resolved. > Must start automatically OOTB to support the tire-kicking, web-start uses. > Must leverage as much 3rd party code as possible to minimize ownership > costs. > Must pass code review. > Unit test code coverage. > Basic config and deployment documentation. > > > What is missing? Anything? Thanks, L > > > > On Jul 17, 2012, at 2:58 PM, Lance Speelmon <la...@rsmart.com> wrote: > > Is there any way to break this work down into chunks? e.g. > > 1. Adopt java PP as default PP moving forward. What are the acceptance > criteria? > 2. Enhance new java PP with message queue abilities. > > WDYT? Thanks, L > > On Jul 17, 2012, at 8:34 AM, Carl Hall <c...@hallwaytech.com> wrote: > > Each app server could run it's own queues but that wouldn't support building > a farm of PP processors unless we also teach them to talk to multiple JMS > servers. Maybe something like DNS round-robin would suffice? > > On Tue, Jul 17, 2012 at 8:25 AM, Erik Froese <erik.fro...@gmail.com> wrote: > > Do we need to cluster activemq? Can't each app server service its own > queues? > Erik > > On Tue, Jul 17, 2012 at 11:23 AM, Carl Hall <c...@hallwaytech.com> wrote: >> What Erik describes has been on the dev wish list for a little while now. >> Moving to an event-driven model would allow us to build out concurrency >> but >> there also comes the question of clustering ActiveMQ. >> >> >> On Thu, Jul 12, 2012 at 6:27 AM, Erik Froese <erik.fro...@gmail.com> >> wrote: >>> >>> Hey David, >>> >>> The code is not clustered. >>> >>> You'd need to write an event listener that would fire when new content >>> is uploaded. It would put the content ids on a JMS queue. Then >>> implement a ContentFetcher that grabs a message off of the queue and >>> wire that into the PPI. Events and Messages are not clustered in OAE >>> (AFAIK) so this would have to be run on each app server. >>> >>> While we're in event-land it'd be nice to be able to regenerate a >>> preview when a content body is updated. I'm not sure if this is >>> possible yet. >>> >>> I'm not sure how we'd limit the CPU usage yet either. You could manage >>> the quartz schedule that runs the PPI. >>> >>> We can also disable concurrent executions of the job. >>> >>> Erik >>> >>> On Wed, Jul 11, 2012 at 8:44 PM, Roma, David <dr...@csu.edu.au> wrote: >>> > Awesome news Erik! >>> > >>> > Our Ops guys will be stoked when we can get this in.. A couple of >>> > questions from someone who hasn't looked at the code or read too >>> > deeply.... >>> > - Does it support clustering >>> > -e.g. can we just run it side by side on each of our app >>> > servers >>> > and they will play nice sharing out processing jobs? >>> > -will it affect performance of the app servers much? Can we >>> > limit the preview processor to say 10%cpu and 500mb ram or low priority >>> > threads or limit the number of items to process or something? This >>> > would >>> > make for a nice simple deployment that wouldn't threaten the app server >>> > stability. >>> > >>> > Cheers, >>> > Dave. >>> > >>> > >>> > >>> > >>> > >>> > -----Original Message----- >>> > From: oae-dev-boun...@collab.sakaiproject.org >>> > [mailto:oae-dev-boun...@collab.sakaiproject.org] On Behalf Of Erik >>> > Froese >>> > Sent: Thursday, 12 July 2012 2:37 AM >>> > To: Carl Hall >>> > Cc: oae-dev@collab.sakaiproject.org; Clay Fenlason >>> > Subject: Re: [oae-dev] Moving the preview processor to java >>> > >>> > Hey everyone, >>> > >>> > Its been a few months but I actually implemented the Java preview >>> > processor as an OSGi bundle. I filed a ticket for it [1] >>> > >>> > I'm not sure where to go from here. Is this something that could be >>> > included POST 1.4.0? >>> > Should I open a PR so we can review the code? If so, PR against which >>> > branch? >>> > >>> > Either way, have a look, give it a go. We'll probably wind up using it >>> > at rSmart. >>> > >>> > Erik >>> > >>> > [1] https://jira.sakaiproject.org/browse/KERN-3021 >>> > >>> > >>> > >>> > On Tue, Apr 17, 2012 at 9:09 AM, Carl Hall <c...@hallwaytech.com> >>> > wrote: >>> >> I totally agree that we should ally ourselves with other communities. >>> >> I >>> >> see >>> >> where we get docsplit from DocumentCloud[1] and we use several other >>> >> libraries for processing that they've most likely contributed to. >>> >> The Java approach is very little custom code compared to the libraries >>> >> we're >>> >> getting from Apache (tika, sanselan, commons, pdfbox), so we would >>> >> still >>> >> building on the shoulders of our friendly community giants. >>> >> >>> >> 1 https://github.com/documentcloud/docsplit >>> >> >>> >> >>> >> >>> >> On Sat, Apr 14, 2012 at 5:43 AM, John Norman <j...@caret.cam.ac.uk> >>> >> wrote: >>> >>> >>> >>> My recollection (perhaps wrong) is that we got this from Document >>> >>> Cloud >>> >>> and I /think/ Chris Roby found it. Document Cloud seems a very >>> >>> relevant and >>> >>> valuable project. If we were able to help them while helping >>> >>> ourselves, >>> >>> other good things could come from the relationship. My general point >>> >>> is that >>> >>> we are thin on resources and so, in principle, symbiotic >>> >>> relationships >>> >>> are >>> >>> helpful. >>> >>> >>> >>> http://www.documentcloud.org/home >>> >>> >>> >>> John >>> >>> >>> >>> Sent from my iPad >>> >>> >>> >>> On 13 Apr 2012, at 17:03, Carl Hall <c...@hallwaytech.com> wrote: >>> >>> >>> >>> I agree with Daniel that our modifications to the preview processor >>> >>> have >>> >>> put its ownership square on us. Was there a community that this >>> >>> script >>> >>> was >>> >>> borrowed from? I thought it was original development that uses >>> >>> various >>> >>> external libraries to do the actual work. This is the approach that >>> >>> Erik is >>> >>> taking with the rewrite using things like Tika (text extraction), >>> >>> Sanselan >>> >>> (images) and a Java port of the python topia.termextract library. >>> >>> >>> >>> I certainly don't deny the speed of development that was realized in >>> >>> creating the PP but the current state of the code is a mess at best. >>> >>> Reuse >>> >>> of libraries in Java is showing a fast rewrite with very little >>> >>> managed code >>> >>> on our part. >>> >>> >>> >>> >>> >>> On Fri, Apr 13, 2012 at 12:50 AM, Daniel Parry >>> >>> <dan...@caret.cam.ac.uk> >>> >>> wrote: >>> >>>> >>> >>>> On Thu, Apr 12, 2012 at 04:21:36PM -0400, Clay Fenlason wrote: >>> >>>> > I think this response is at best orthogonal to the point John's >>> >>>> > trying >>> >>>> > to raise, though I gather this kind of reaction must come from a >>> >>>> > buildup of some real frustration around the PP, which I don't mean >>> >>>> > to >>> >>>> > discount. I also think John was pretty clear about what he was >>> >>>> > suggesting: that there be a conversation with the community we got >>> >>>> > the >>> >>>> > PP from, if the conversation hasn't happened already, to see if >>> >>>> > there >>> >>>> > might still be a way to work together before we decide to just own >>> >>>> > it >>> >>>> > ourselves. >>> >>>> >>> >>>> I'd suggest the way that the preview processor was being extended >>> >>>> (initially a >>> >>>> python server add on, followed by a ruby rewrite for tag extraction) >>> >>>> and >>> >>>> the >>> >>>> variety of ruby versions that deployers were using and the methods >>> >>>> used >>> >>>> to >>> >>>> deploy it were indicative of a) the OAE community already 'owning' >>> >>>> the PP >>> >>>> and b) >>> >>>> as has already been pointed out some standardization needed >>> >>>> restoring >>> >>>> and >>> >>>> additional functionality added for deployers. Hence, the list was >>> >>>> pinged[0] a >>> >>>> while back to ask about standardizing and extending in java. I'm not >>> >>>> sure >>> >>>> of any >>> >>>> other way to contact the original PP community or if such a >>> >>>> community >>> >>>> separate >>> >>>> to OAE even still exists? >>> >>>> >>> >>>> Best wishes, >>> >>>> >>> >>>> Daniel >>> >>>> >>> >>>> [0] >>> >>>> >>> >>>> >>> >>>> http://collab.sakaiproject.org/pipermail/oae-dev/2012-April/001677.html >>> >>>> >>> >>>> -- >>> >>>> --| Daniel Parry: dan...@caret.cam.ac.uk. www.caret.cam.ac.uk/ |-- >>> >>>> "Of all the things a leader should fear, complacency should >>> >>>> head the list." [John C. Maxwell] >>> >>>> _______________________________________________ >>> >>>> oae-dev mailing list >>> >>>> oae-dev@collab.sakaiproject.org >>> >>>> http://collab.sakaiproject.org/mailman/listinfo/oae-dev >>> >>> >>> >>> >>> >>> _______________________________________________ >>> >>> oae-dev mailing list >>> >>> oae-dev@collab.sakaiproject.org >>> >>> http://collab.sakaiproject.org/mailman/listinfo/oae-dev >>> >> >>> >> >>> >> >>> >> _______________________________________________ >>> >> oae-dev mailing list >>> >> oae-dev@collab.sakaiproject.org >>> >> http://collab.sakaiproject.org/mailman/listinfo/oae-dev >>> >> >>> > _______________________________________________ >>> > oae-dev mailing list >>> > oae-dev@collab.sakaiproject.org >>> > http://collab.sakaiproject.org/mailman/listinfo/oae-dev >>> > Charles Sturt University >>> > >>> > | ALBURY-WODONGA | BATHURST | CANBERRA | DUBBO | GOULBURN | MELBOURNE | >>> > ONTARIO | ORANGE | PORT MACQUARIE | SYDNEY | WAGGA WAGGA | >>> > >>> > LEGAL NOTICE >>> > This email (and any attachment) is confidential and is intended for the >>> > use of the addressee(s) only. If you are not the intended recipient of >>> > this >>> > email, you must not copy, distribute, take any action in reliance on it >>> > or >>> > disclose it to anyone. Any confidentiality is not waived or lost by >>> > reason >>> > of mistaken delivery. Email should be checked for viruses and defects >>> > before >>> > opening. Charles Sturt University (CSU) does not accept liability for >>> > viruses or any consequence which arise as a result of this email >>> > transmission. Email communications with CSU may be subject to automated >>> > email filtering, which could result in the delay or deletion of a >>> > legitimate >>> > email before it is read at CSU. The views expressed in this email are >>> > not >>> > necessarily those of CSU. >>> > >>> > Charles Sturt University in Australia http://www.csu.edu.au The >>> > Chancellery, Panorama Avenue, Bathurst NSW Australia 2795 ABN: 83 878 >>> > 708 >>> > 551; CRICOS Provider Numbers: 00005F (NSW), 01947G (VIC), 02960B (ACT) >>> > >>> > Charles Sturt University in Ontario http://www.charlessturt.ca 860 >>> > Harrington Court, Burlington Ontario Canada L7N 3N4 Registration: >>> > www.peqab.ca >>> > >>> > Consider the environment before printing this email. >> >> > > > _______________________________________________ > oae-dev mailing list > oae-dev@collab.sakaiproject.org > http://collab.sakaiproject.org/mailman/listinfo/oae-dev > > > > _______________________________________________ > oae-dev mailing list > oae-dev@collab.sakaiproject.org > http://collab.sakaiproject.org/mailman/listinfo/oae-dev > > > > _______________________________________________ > oae-dev mailing list > oae-dev@collab.sakaiproject.org > http://collab.sakaiproject.org/mailman/listinfo/oae-dev > > > _______________________________________________ > oae-dev mailing list > oae-dev@collab.sakaiproject.org > http://collab.sakaiproject.org/mailman/listinfo/oae-dev > > > _______________________________________________ > oae-dev mailing list > oae-dev@collab.sakaiproject.org > http://collab.sakaiproject.org/mailman/listinfo/oae-dev > > > > _______________________________________________ > oae-dev mailing list > oae-dev@collab.sakaiproject.org > http://collab.sakaiproject.org/mailman/listinfo/oae-dev > _______________________________________________ oae-dev mailing list oae-dev@collab.sakaiproject.org http://collab.sakaiproject.org/mailman/listinfo/oae-dev
