Hi Karl, I want to put the last version of Tika (1.16) into MCF 2.8.1. I already did it with MCF 2.6 and Tika 1.15 but not in a clean way since I have put all the Tika libs in the ‘lib’ and ‘connector-common-lib’ folders of MCF, removing the MCF libs that where in an older version. Despite it was not clean, it worked. This time I want to properly do the things but I don’t really understand how the libs are splitted between the folders ‘lib’, ‘connector-common-lib’ and ‘connector-lib’, knowing that the Tika libs are spread between them. I have to be careful cause for example I noticed that MCF 2.8.1 uses the 21 version of the guava lib where Tika uses the version 17, but there are some major changes since the version 19 that can potentially break something in Tika. Can you help me a little bit with this please ?
Thanks, Julien De : Karl Wright Envoyé le :vendredi 8 septembre 2017 21:43 À : user@manifoldcf.apache.org Objet :Re: Question about ManifoldCF 2.8 Hi Othman, There are two properties files for zookeeper: the global properties, and the local (zookeeper managed) properties. The database configuration is in the zookeeper managed properties. Please examine the following page for setting up Postgresql properties: https://manifoldcf.apache.org/release/release-2.8.1/en_US/how-to-build-and-deploy.html Indexable files are files that those that the output connector says can be indexed. It's a function of the output connector and its configuration. Thanks, Karl On Fri, Sep 8, 2017 at 2:07 PM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Sorry to bother you again, but what is the difference between indexable files and files in the path tab of a job ? Thanks, Othman BELHAJ On Fri, 8 Sep 2017 at 19:27, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, My zookeeper is still pointing to the HSQL database. What should I do in order to change it so that it points to my PostgreSQL database ? Best regards, Othman Belhaj . On Wed, 6 Sep 2017 at 15:34, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Thank you, Karl. I will try to combine Postgresql with zookeeper and let you know. Othman. On Wed, 6 Sep 2017 at 13:18, Karl Wright <daddy...@gmail.com> wrote: No, you can use whatever supported database you like. Karl On Wed, Sep 6, 2017 at 6:58 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: As far as I know, when you use zookeeper , you obligatory need to use HSQLDB to go with it, right? Thanks, Othman On Wed, 6 Sep 2017 at 12:56, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, HSQLDB stores all tables in memory so you need to size it accordingly. That is one reason we prefer Postgresql for production deployments. Thanks, Karl On Wed, Sep 6, 2017 at 6:21 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, I resolved the elasticsearch problem however the application doesn't seem to work after I have run a job to crawl over 500k documents. I get an GC overhead limit exceeded in the hsql database. How many should I allocate for it? Best regards, Othman On Tue, 5 Sep 2017 at 12:43, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, Thanks for doing the evaluation of the problem. Generally, the ManifoldCF project does not have the expertise to diagnose problems with external systems like Solr or Elasticsearch. So going to another newsgroup for those kinds of issues would be a good idea. Thanks! Karl On Tue, Sep 5, 2017 at 4:33 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, I have analyzed the error and found out that it was mainly an elasticsearch problem. I saw in some forums that one of the adopted solution is to modify elasticsearch.yml and set the http.max_content_length to a greater value. However, the job got stuck in the last two indexable files ( two pptx files with 22Mo and 2Mo respectively). The job eventually ended but a stack trace showed that elasticsearch ran out of memory. For your information, I have allocated 4Go for elasticsearch execution. Is it enough in order to have a good performance. You will find attached the stack traces of elasticsearch. Best regards, Othman BELHAJ. On Mon, 4 Sep 2017 at 16:40, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, I'm sorry to bother on your holiday. I will try to analyze it today and let it you know what I have found. Enjoy your day ! Best regards, Othman BELHAJ. On Mon, 4 Sep 2017 at 16:06, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, I won't be able to look at this today; it is a holiday here. But, the "socket write" error is coming from ElasticSearch. If ES is configured to not accept documents greater than a certain size, that might explain it. Maybe the ES logs would help? I'm afraid you're going to need to do the work to find out what is going wrong in those cases now. Thanks, Karl On Mon, Sep 4, 2017 at 4:53 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, This morning, I have tried the zookeeper based file and it worked really good. However, I still have one error which is bugging me. It is a socket write error. You will find attached the simple history report. Surprisingly, I didn't have any stack trace in the ManifoldCF log file. Best regards, Othman. On Fri, 1 Sep 2017 at 19:39, Karl Wright <daddy...@gmail.com> wrote: This is from file locking yet again. I have uploaded a new RC. Please download and try out the zookeeper locking. https://dist.apache.org/repos/dist/dev/manifoldcf/apache-manifoldcf-2.8.1 Karl On Fri, Sep 1, 2017 at 1:11 PM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: There is another issue as well that gives the following stack trace. Othman. On Fri, 1 Sep 2017 at 18:05, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, I took the binary from the ManifoldCF 2.8.1 RC0. It had the version 3.9 of POI and when I changed the version to 3.15 it worked fine. I really want to try the zookeeper if as you told me its performance is better than the file-based example. For the time being, I'm using the file-based because it is the only part that works for me but I actually need a stable version for my production environment. That is one point. Another point is, the path's tab is still an issue for me because I exclude some files and it still crawls them. I want to exclude some specific extensions of files and some specific directories. For instance, i don't want to index .exe files and contains a specific word. I do as follows I make the first exclude with *.exe and the second one with *word*. Only the second one which doesn't work. How can I solve this issue, please? Thank you very much, have a nice week-end, Othman On Fri, 1 Sep 2017 at 16:46, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, I will respin a new 2.8.1 (RC1) to address the zookeeper issue. The failure you are seeing is "NoSuchMethodError". Therefore, the class is being found, but it is the *wrong* class. When you deployed the new release, did you deploy it in a new directory, or did you overwrite the previous deployment? If you overwrote it, you probably have multiple versions of the POI jars. Karl On Fri, Sep 1, 2017 at 9:59 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, I have just tried the new release of ManifoldCF. At first, the first job ended normally, but in the second I got a new stack trace concerning the POI. Moreover, the runzookeeper.bat doesn't run properly. It shows me the stack trace attached. Ps: The second attached file contains the POI stack trace. Othman. On Fri, 1 Sep 2017 at 12:21, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, You do not need a new database instance. You can download MCF 2.8.1 RC0 from here: https://dist.apache.org/repos/dist/dev/manifoldcf/apache-manifoldcf-2.8.1 Karl On Fri, Sep 1, 2017 at 5:42 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, Thank you very much for your help, I'm going to try out the zookeeper example. Should I initialize a new database? And how can I run the zookeeper start-agent ? Othman. On Fri, 1 Sep 2017 at 11:37, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, These exceptions are now coming from file locking and are due to permissions problems. I suggest you go to Zookeeper for file locking. I am building a 2.8.1 release candidate. When it available for download, I'll send you the URL. Thanks, Karl On Fri, Sep 1, 2017 at 5:27 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, This morning, I have followed the steps you told me to do and I still got stack traces. I have attached the stack traces as well as the content of my lib repo and option.env. I have installed zookeeper and I'm ready to use the zookeeper example. Could you guide through it? I don't know if I follow the same steps in the file based example, I may not get stack traces. Thanks, Othman On Thu, 31 Aug 2017 at 18:19, Karl Wright <daddy...@gmail.com> wrote: Please do the following: (0) Shut down all ManifoldCF processes. (1) Move poi*.jar from connector-common-lib to lib. (2) Move dom4j*.jar from connector-common-lib to lib. (3) Move commons-collections4*.jar from connector-common-lib to lib. (4) Move xmlbeans*.java from connector-common-lib to lib. (5) Move curvesapi*.jar from connector-common-lib to lib. (6) Modify your options.env to include all of the jars you moved. (7) Start up all ManifoldCF processes. (8) If you still get stack traces, please send them to me. Karl On Thu, Aug 31, 2017 at 12:12 PM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, By 'other place', do you mean the \lib repository? If that so, then I have already tried it and it didn't work. Othman. On Thu, 31 Aug 2017 at 18:07, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, I used the java dependency inspector to see what the issue is and it turns out that poi-ooxml.jar does refer back to poi.jar in the class that is failing. So you will need to move poi-3.15.jar and commons-collections4-1.4.jar to the other place as well. Let's hope that finally fixes this issue. I'm very unhappy about the quality of the POI project code; it is definitely not using reasonable engineering practices, and I will be opening a ticket with them. Thanks, Karl On Thu, Aug 31, 2017 at 11:57 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: I'm using the file based example and all the changes you told me to do. I reproduced them in the file based example. I'll try to install zookeeper and use the zookeeper example. Will I need a configuration to do in order to run the zookeeper example ? Othman. On Thu, 31 Aug 2017 at 17:46, Karl Wright <daddy...@gmail.com> wrote: Are you using the zookeeper example, or the file-based example? If these jars have all been moved, and the options.env includes them, then I have to conclude that Apache POI's pom.xml is incorrect too. It will take a while to figure out what's missing that poi-ooxml.jar needs that is not listed. Karl On Thu, Aug 31, 2017 at 11:39 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: All the dependencies you mentioned have already been added in the options.env.win file in the multiprocess-file-example repository. On Thu, 31 Aug 2017 at 17:33, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Yes, I added it in the options.env.win file. Should it be the one in the multiprocess-zk-example document or multiprocess-file-example ? On Thu, 31 Aug 2017 at 17:30, Karl Wright <daddy...@gmail.com> wrote: It's not related at all to elasticsearch. Karl On Thu, Aug 31, 2017 at 11:26 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Could it be a problem of elasticsearch's version ? I'm actually using 2.1.0 which is pretty old for this new version of ManifoldCF? Othman. On Thu, 31 Aug 2017 at 17:23, Beelz Ryuzaki <i93oth...@gmail.com> wrote: I moved back both the jars you mentioned and a different is showing. You will find the stack trace attached. Thanks, Othman On Thu, 31 Aug 2017 at 17:09, Karl Wright <daddy...@gmail.com> wrote: I've looked at the dependencies; you should not have moved poi-3.15.jar. Please move that back, and commons-collections4-4.1.jar too. You *will* need to move curvesapi-1.04.jar though. Thanks, Karl On Thu, Aug 31, 2017 at 11:04 AM, Karl Wright <daddy...@gmail.com> wrote: If you include poi.jar, then all dependencies of poi.jar must also be included. This would mean that curvesapi-1.04.jar and commons-collections4-4.1.jar should also be included. Karl On Thu, Aug 31, 2017 at 10:23 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Karl, I added the two jars that you have mentioned and another one : poi-3.15.jar . Unfortunately, there is another error showing. This time, it concerns excel files. You will find attached the stack trace. Othman. On Thu, 31 Aug 2017 at 15:32, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, Yes, this shows that the jar we moved calls back into another jar, which will also need to be moved. *That* jar has yet another dependency too. The list of jars is thus extended to include: poi-ooxml-3.15.jar dom4j-1.6.1.jar Karl On Thu, Aug 31, 2017 at 9:25 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: You will find attached the stack trace. My apologies for the bad quality of the image, I'm doing my best to send you the stack trace as I don't have the right to send documents outside the company. Thank you for your time, Othman On Thu, 31 Aug 2017 at 15:16, Karl Wright <daddy...@gmail.com> wrote: Once again, I need a stack trace to diagnose what the problem is. Thanks, Karl On Thu, Aug 31, 2017 at 9:14 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Oh, actually it didn't solve the problem. I looked into the log file and saw the following error: Error tossed : org/apache/poi/POIXMLTypeLoader java.lang.NoClassDefFoundError: org/apache/poi/POIXMLTypeLoader. Maybe another jar is missing ? Othman. On Thu, 31 Aug 2017 at 15:01, Beelz Ryuzaki <i93oth...@gmail.com> wrote: I have tried what you told me to do, and you expected the crawling resumed. How about the regular expressions? How can I make complex regular expressions in the job's paths tab ? Thank you very much for your help. Othman. On Thu, 31 Aug 2017 at 14:47, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Ok, I will try it right away and let you know if it works. Othman. On Thu, 31 Aug 2017 at 14:15, Karl Wright <daddy...@gmail.com> wrote: Oh, and you also may need to edit your options.env files to include them in the classpath for startup. Karl On Thu, Aug 31, 2017 at 7:53 AM, Karl Wright <daddy...@gmail.com> wrote: If you are amenable, there is another workaround you could try. Specifically: (1) Shut down all MCF processes. (2) Move the following two files from connector-common-lib to lib: xmlbeans-2.6.0.jar poi-ooxml-schemas-3.15.jar (3) Restart everything and see if your crawl resumes. Please let me know what happens. Karl On Thu, Aug 31, 2017 at 7:33 AM, Karl Wright <daddy...@gmail.com> wrote: I created a ticket for this: CONNECTORS-1450. One simple workaround is to use the external Tika server transformer rather than the embedded Tika Extractor. I'm still looking into why the jar is not being found. Karl On Thu, Aug 31, 2017 at 7:08 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Yes, I'm actually using the latest binary version, and my job got stuck on that specific file. The job status is still Running. You can see it in the attached file. For your information, the job started yesterday. Thanks, Othman On Thu, 31 Aug 2017 at 13:04, Karl Wright <daddy...@gmail.com> wrote: It looks like a dependency of Apache POI is missing. I think we will need a ticket to address this, if you are indeed using the binary distribution. Thanks! Karl On Thu, Aug 31, 2017 at 6:57 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: I'm actually using the binary version. For security reasons, I can't send any files from my computer. I have copied the stack trace and scanned it with my cellphone. I hope it will be helpful. Meanwhile, I have read the documentation about how to restrict the crawling and I don't think the '|' works in the specified. For instance, I would like to restrict the crawling for the documents that counts the 'sound' word . I proceed as follows: *(SON)* . the document is with capital letters and I noticed that it didn't take it into consideration. Thanks, Othman On Thu, 31 Aug 2017 at 12:40, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, The way you restrict documents with the windows share connector is by specifying information on the "Paths" tab in jobs that crawl windows shares. There is end-user documentation both online and distributed with all binary distributions that describe how to do this. Have you found it? Karl On Thu, Aug 31, 2017 at 5:25 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hello Karl, Thank you for your response, I will start using zookeeper and I will let you know if it works. I have another question to ask. Actually, I need to make some filters while crawling. I don't want to crawl some files and some folders. Could you give me an example of how to use the regex. Does the regex allow to use /i to ignore cases ? Thanks, Othman On Wed, 30 Aug 2017 at 19:53, Karl Wright <daddy...@gmail.com> wrote: Hi Beelz, File-based sync is deprecated because people often have problems with getting file permissions right, and they do not understand how to shut processes down cleanly, and zookeeper is resilient against that. I highly recommend using zookeeper sync. ManifoldCF is engineered to not put files into memory so you do not need huge amounts of memory. The default values are more than enough for 35,000 files, which is a pretty small job for ManifoldCF. Thanks, Karl On Wed, Aug 30, 2017 at 11:58 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: I'm actually not using zookeeper. i want to know how is zookeeper different from file based sync? I also need a guidance on how to manage my pc's memory. How many Go should I allocate for the start-agent of ManifoldCF? Is 4Go enough in order to crawler 35K files ? Othman. On Wed, 30 Aug 2017 at 16:11, Karl Wright <daddy...@gmail.com> wrote: Your disk is not writable for some reason, and that's interfering with ManifoldCF 2.8 locking. I would suggest two things: (1) Use Zookeeper for sync instead of file-based sync. (2) Have a look if you still get failures after that. Thanks, Karl On Wed, Aug 30, 2017 at 9:37 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hi Mr Karl, Thank you Mr Karl for your quick response. I have looked into the ManifoldCF log file and extracted the following warnings : - Attempt to set file lock 'D:\xxxx\apache_manifoldcf-2.8\multiprocess-file-example\.\.\synch area\569\352\lock-_POOLTARGET_OUTPUTCONNECTORPOOL_ES (Lowercase) Synapses.lock' failed : Access is denied. - Couldn't write to lock file; disk may be full. Shutting down process; locks may be left dangling. You must cleanup before restarting. ES (lowercase) synapses being the elasticsearch output connection. Moreover, the job uses Tika to extract metadata and a file system as a repository connection. During the job, I don't extract the content of the documents. I was wandering if the issue comes from elasticsearch ? Othman. On Wed, 30 Aug 2017 at 14:08, Karl Wright <daddy...@gmail.com> wrote: Hi Othman, ManifoldCF aborts a job if there's an error that looks like it might go away on retry, but does not. It can be either on the repository side or on the output side. If you look at the Simple History in the UI, or at the manifoldcf.log file, you should be able to get a better sense of what went wrong. Without further information, I can't say any more. Thanks, Karl On Wed, Aug 30, 2017 at 5:33 AM, Beelz Ryuzaki <i93oth...@gmail.com> wrote: Hello, I'm Othman Belhaj, a software engineer from société générale in France. I'm actually using your recent version of manifoldCF 2.8 . I'm working on an internal search engine. For this reason, I'm using manifoldcf in order to index documents on windows shares. I encountered a serious problem while crawling 35K documents. Most of the time, when manifoldcf start crawling a big sized documents (19Mo for example), it ends the job with the following error: repeated service interruptions - failure processing document : software caused connection abort: socket write error. Can you give me some tips on how to solve this problem, please ? I use PostgreSQL 9.3.x and elasticsearch 2.1.0 . I'm looking forward for your response. Best regards, Othman BELHAJ --- L'absence de virus dans ce courrier électronique a été vérifiée par le logiciel antivirus Avast. https://www.avast.com/antivirus
