Hello I thought you would like to know that we solved this issue. I've analysed SaxonServlet sources and found that some relevant bit are synchronised and HttpClient use safe multithreaded connection manager. When I've tried to reconfigure some of manger's parameters and test them under a load I discovered that every Saxon Servlet Logs has accompanying error logs in the Fedore server ... java.io.IOException: Bad file descriptor ...
As it happens we are currently on Fedora 3.4.1 and above behaviour is known bug reported here https://jira.duraspace.org/browse/FCREPO-774. Basically SaxonServlet couldn't complete transformation as the XML stream were incomplete due to wrong garbage collection. The solution appears to be updating fcrepo-server-*.jar form Fedora 3.4.2 release as suggested here http://fedora-commons.1317035.n2.nabble.com/fcrepo-user-Occasional-random-FilterSetup-Can-t-do-next-doFilter-error-td5657344.html. Sebastian == Sebastian Pałucha IS Specialist (Main Library) Durham University Computing and Information Services ________________________________________ Date: Mon, 27 Feb 2012 17:40:31 +0000 From: "James, Eric" <[email protected]> Subject: Re: [fcrepo-user] Still having trouble with CMA content models ... To: "Support and info exchange list for Fedora users." <[email protected]> Richard, I have not had an problems with SaxonServlet, but did run into a similar issue when trying to implement my own custom transformer. The symptoms were that small XML sources worked, but larger ones sometimes failed. The problem was concurrent use of the same servlet instance trying to parse the XML, resulting in mashed XML - something that was more likely to happen with big XMLs taking a long time to process. The solution was to synchronize the method within the servlet, so one transformation was guaranteed to end before another began. -Eric ________________________________________ From: [email protected] [[email protected]] Sent: Monday, February 27, 2012 9:02 AM To: Support and info exchange list for Fedora users. Subject: Re: [fcrepo-user] Still having trouble with CMA content models ... -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I'm afraid I'm not familiar with SaxonServlet enough to answer your question directly, but here's some friendly advice: if you're going to be moving 2.5MB XML files through dynamic disseminations on a regular basis, you may want to look to a more powerful backend service than SaxonServlet. To my knowledge, SaxonServlet doesn't provide any content caching or optimize the provision of resources for transforms. The old go-to for a lot of folks in this situation used to be Apache Cocoon, but I know there are plenty of nice choices for XML pipelining out there these days. Incidentally, this is one reason why we (UVa) chose years ago for what used to be called "atomistic" representations of archives, i.e. the disassembly of complex archival contexts into object graphs. The processing for retrieval you are doing as XSLT we do as SPARQL against the Resource Index, which is optimized for just this kind of graph traversal. - --- A. Soroka Software & Systems Engineering :: Online Library Environment the University of Virginia Library On Feb 27, 2012, at 7:58 AM, HIGGINS R.I. wrote: > Hello: > Still working with these, trying to extract a single item description > from an EAD catalogue. > > With a relatively small EAD file (ca. 100KB) this works fine: > > http://endure.dur.ac.uk:8080/saxon/SaxonServlet?source=http://endure.d > ur > .ac.uk:8080/fedora/objects/UkDhU:EADCatalogue.0162/datastreams/EAD/con > te > nt&style=http://endure.dur.ac.uk:8080/fedora/objects/UkDhU:Styles.0011 > /d > atastreams/XSL/content&clear-stylesheet-cache=yes&EADCatNo=UkDhU:EADCa > ta > logue.0162&RefID1=ORD-11&FedoraPID=UkDhU:DCD_Misc.Ch.3650 > > However, a larger EAD file (ca. 2.5MB) only works every other time > > http://endure.dur.ac.uk:8080/saxon/SaxonServlet?source=http://endure.d > ur > .ac.uk:8080/fedora/objects/UkDhU:EADCatalogue.0090/datastreams/EAD/con > te > nt&style=http://endure.dur.ac.uk:8080/fedora/objects/UkDhU:Styles.0011 > /d > atastreams/XSL/content&clear-stylesheet-cache=yes&EADCatNo=UkDhU:EADCa > ta > logue.0090&RefID1=MC-3650&FedoraPID=UkDhU:DCD_Misc.Ch.3650 > > alternating with returning an HTTP 500 message > org.xml.sax.SAXParseException: XML document structures must start and > end within the same entity, which every now and then adds a specific > EAD element mismatch to the message. > > The above URLs are external versions of what the CMA pushes around and > in this context only the RefID1 parameter of the three is used. > > Both work every time with a standalone copy of Saxon (and as a minor > secondary point the XML output from that picks up the css file and > formats correctly too). > >> From the nature of the errors I would guess that this is either a > threading or a capacity issue with the fedora Saxon servlet (using > version 3.4.1 here)? > Is there any setting that can be adjusted to fix this? > As ever, baffled, but grateful for any ideas. > - - - - - > # Richard Higgins > # Durham University Library > # Archives & Special Collections > # Palace Green > # Durham > # DH1 3RN > # E-Mail: [email protected] > > > ---------------------------------------------------------------------- > -------- Try before you buy = See our experts in action! > The most comprehensive online learning library for Microsoft > developers is just $99.99! Visual Studio, SharePoint, SQL - plus > HTML5, CSS3, MVC3, Metro Style Apps, more. Free future releases when you > subscribe now! > http://p.sf.net/sfu/learndevnow-dev2 > _______________________________________________ > Fedora-commons-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/fedora-commons-users -----BEGIN PGP SIGNATURE----- Version: GnuPG/MacGPG2 v2.0.17 (Darwin) Comment: GPGTools - http://gpgtools.org iQEcBAEBAgAGBQJPS40MAAoJEATpPYSyaoIkU7EIAJ7jJA3NUKhFpHqjysLoYbdW L8yVIqiKNDths/6zeFYMaIghyHM2F8ZXhxT7zGfu76/sGzvBl6ibvsm64lmObtGj bAI5vtsXo1Mk6x5dQT5ipvw99N4o3IBPlaVox6lDKxspjn84HdIGEfPeA16GoVT6 klyFYy69SdgcUWVsePxQu/gQxN3XO8m82opSrEZMeuE5tbMK+h+6q+2PzSBfUoiU 4OiBf4xzoHbZIVkvrmFft9678XiGwMxxGeu4euSmahNLeTuVXNwO+aS12UTmI9rv x+JLo9WoXeEBoKoe2XI1pMkh+8PgoXylY0uROAipEecPueucaH4zQXUwU0k7VAY= =7/0U -----END PGP SIGNATURE----- ------------------------------------------------------------------------------ Try before you buy = See our experts in action! The most comprehensive online learning library for Microsoft developers is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3, Metro Style Apps, more. Free future releases when you subscribe now! http://p.sf.net/sfu/learndevnow-dev2 _______________________________________________ Fedora-commons-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fedora-commons-users ------------------------------ ------------------------------------------------------------------------------ Try before you buy = See our experts in action! The most comprehensive online learning library for Microsoft developers is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3, Metro Style Apps, more. Free future releases when you subscribe now! http://p.sf.net/sfu/learndevnow-dev2 ------------------------------ _______________________________________________ Fedora-commons-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fedora-commons-users End of Fedora-commons-users Digest, Vol 60, Issue 22 **************************************************** ------------------------------------------------------------------------------ This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure _______________________________________________ Fedora-commons-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fedora-commons-users
