I checked IITA’s 259 Feb 14 records from last month for duplicates using Atmire’s Duplicate Checker on a fresh snapshot of CGSpace on my local machine and everything looks good
I am now only waiting to hear from her about where the items should go, though I assume Journal Articles go to IITA Journal Articles collection, etc…
Looking at the other half of Udana’s WLE records from 2018-11
I finished the ones for Restoring Degraded Landscapes (RDL), but these are for Variability, Risks and Competing Uses (VRC)
I did the usual cleanups for whitespace, added regions where they made sense for certain countries, cleaned up the DOI link formats, added rights information based on the publications page for a few items
Most worryingly, there are encoding errors in the abstracts for eleven items, for example:
68.15% <20> 9.45 instead of 68.15% ± 9.45
2003<EFBFBD>2013 instead of 2003–2013
I think I will need to ask Udana to re-copy and paste the abstracts with more care using Google Docs
I checked IITA’s 259 Feb 14 records from last month for duplicates using Atmire’s Duplicate Checker on a fresh snapshot of CGSpace on my local machine and everything looks good
I am now only waiting to hear from her about where the items should go, though I assume Journal Articles go to IITA Journal Articles collection, etc…
Looking at the other half of Udana’s WLE records from 2018-11
I finished the ones for Restoring Degraded Landscapes (RDL), but these are for Variability, Risks and Competing Uses (VRC)
I did the usual cleanups for whitespace, added regions where they made sense for certain countries, cleaned up the DOI link formats, added rights information based on the publications page for a few items
Most worryingly, there are encoding errors in the abstracts for eleven items, for example:
68.15% <20> 9.45 instead of 68.15% ± 9.45
2003<EFBFBD>2013 instead of 2003–2013
I think I will need to ask Udana to re-copy and paste the abstracts with more care using Google Docs
<li>I checked IITA’s 259 Feb 14 records from last month for duplicates using Atmire’s Duplicate Checker on a fresh snapshot of CGSpace on my local machine and everything looks good</li>
<li>I am now only waiting to hear from her about where the items should go, though I assume Journal Articles go to IITA Journal Articles collection, etc…</li>
<li>Looking at the other half of Udana’s WLE records from 2018-11
<ul>
<li>I finished the ones for Restoring Degraded Landscapes (RDL), but these are for Variability, Risks and Competing Uses (VRC)</li>
<li>I did the usual cleanups for whitespace, added regions where they made sense for certain countries, cleaned up the DOI link formats, added rights information based on the publications page for a few items</li>
<li>Most worryingly, there are encoding errors in the abstracts for eleven items, for example:</li>
<li>68.15% <20> 9.45 instead of 68.15% ± 9.45</li>
<li>2003<EFBFBD>2013 instead of 2003–2013</li>
</ul></li>
<li>I think I will need to ask Udana to re-copy and paste the abstracts with more care using Google Docs</li>
<li>As I was inspecting the archive I noticed that there were some problems with the bitsreams:
<ul>
<li>First, Sisay didn’t include the bitstream descriptions</li>
<li>Second, only five items had bitstreams and I remember in the discussion with IITA that there should have been nine!</li>
<li>I had to refer to the original CSV from January to find the file names, then download and add them to the export contents manually!</li>
</ul></li>
<li>After adding the missing bitstreams and descriptions manually I tested them again locally, then imported them to a temporary collection on CGSpace:</li>
<li>DSpace’s export function doesn’t include the collections for some reason, so you need to import them somewhere first, then export the collection metadata and re-map the items to proper owning collections based on their types using OpenRefine or something</li>
<li>After re-importing to CGSpace to apply the mappings, I deleted the collection on DSpace Test and ran the <code>dspace cleanup</code> script</li>
<li>Merge the IITA research theme changes from last month to the <code>5_x-prod</code> branch (<ahref="https://github.com/ilri/DSpace/pull/413">#413</a>)
<ul>
<li>I will deploy to CGSpace soon and then think about how to batch tag all IITA’s existing items with this metadata</li>
<li>Generate a controlled vocabulary of 1187 AGROVOC subjects from the top 1500 that I checked last month, dumping the terms themselves using <code>csvcut</code> and then applying XML controlled vocabulary format in vim and then checking with tidy for good measure:</li>
</ul>
<pre><code>$ csvcut -c name 2019-02-22-subjects.csv > dspace/config/controlled-vocabularies/dc-contributor-author.xml
<li>Atmire noticed my message about the “solr_update_time_stamp” error on the dspace-tech mailing list and created an issue on their tracker to discuss it with me
<ul>
<li>They say the error is harmless, but has nevertheless been fixed in their newer module versions</li>
<li><del>Interestingly, if I check an item in the REST API it is also mostly blank: only the title and the ID!</del> On second thought I realize I probably was just seeing the default view without any “expands”</li>
<li>I don’t see anything unusual in the Tomcat logs, though there are thousands of those <code>solr_update_time_stamp</code> errors:</li>
</ul></li>
</ul>
<pre><code># journalctl -u tomcat7 | grep -c 'Multiple update components target the same field:solr_update_time_stamp'
1076
</code></pre>
<ul>
<li>I restarted Tomcat and it’s OK now…</li>
<li>Skype meeting with Peter and Abenet and Sisay
<ul>
<li>We want to try to crowd source the correction of invalid AGROVOC terms starting with the ~313 invalid ones from our top 1500</li>
<li>We will share a Google Docs spreadsheet with the partners and ask them to mark the deletions and corrections</li>
<li>Abenet and Alan to spend some time identifying correct DCTERMS fields to move to, with preference over CG Core 2.0 as we want to be globally compliant (use information from SEO crosswalks)</li>
<li>I need to follow up on the privacy page that Sisay worked on</li>
<li>We want to try to migrate the 600 <ahref="https://livestock.cgiar.org">Livestock CRP blog posts</a> to CGSpace, Peter will try to export the XML from WordPress so I can try to parse it with a script</li>
<li>I shared a post on Yammer informing our editors to try to AGROVOC controlled list</li>
<li>The SPDX legal committee had a meeting and discussed the addition of CC-BY-ND-3.0-IGO and other IGO licenses to their list, but it seems unlikely (<ahref="https://github.com/spdx/license-list-XML/issues/767#issuecomment-470709673">spdx/license-list-XML/issues/767</a>)</li>
<li>The FireOak report highlights the fact that several CGSpace collections have mixed-content errors due to the use of HTTP links in the Feedburner forms</li>
<li>I see 46 occurrences of these with this query:</li>
</ul>
<pre><code>dspace=# SELECT text_value FROM metadatavalue WHERE resource_type_id in (3,4) AND (text_value LIKE '%http://feedburner.%' OR text_value LIKE '%http://feeds.feedburner.%');
</code></pre>
<ul>
<li>I can replace these globally using the following SQL:</li>
</ul>
<pre><code>dspace=# UPDATE metadatavalue SET text_value = REGEXP_REPLACE(text_value, 'http://feedburner.','https//feedburner.', 'g') WHERE resource_type_id in (3,4) AND text_value LIKE '%http://feedburner.%';
UPDATE 43
dspace=# UPDATE metadatavalue SET text_value = REGEXP_REPLACE(text_value, 'http://feeds.feedburner.','https//feeds.feedburner.', 'g') WHERE resource_type_id in (3,4) AND text_value LIKE '%http://feeds.feedburner.%';
UPDATE 44
</code></pre>
<ul>
<li>I ran the corrections on CGSpace and DSpace Test</li>
<li>Working on tagging IITA’s items with their new research theme (<code>cg.identifier.iitatheme</code>) based on their existing IITA subjects (see <ahref="/cgspace-notes/2018-02/">notes from 2019-02</a>)</li>
<li>I exported the entire IITA community from CGSpace and then used <code>csvcut</code> to extract only the needed fields:</li>
<li><p>After importing to OpenRefine I realized that tagging items based on their subjects is tricky because of the row/record mode of OpenRefine when you split the multi-value cells as well as the fact that some items might need to be tagged twice (thus needing a <code>||</code>)</p></li>
<li><p>I think it might actually be easier to filter by IITA subject, then by IITA theme (if needed), and then do transformations with some conditional values in GREL expressions like:</p></li>
</ul>
<pre><code>if(isBlank(value), 'PLANT PRODUCTION & HEALTH', value + '||PLANT PRODUCTION & HEALTH')
</code></pre>
<ul>
<li>Then it’s more annoying because there are four IITA subject columns…</li>
<li>In total this would add research themes to 1,755 items</li>
<li>I want to double check one last time with Bosede that they would like to do this, because I also see that this will tag a few hundred items from the 1970s and 1980s</li>
<li>This is a bit ugly, but it works (using the <ahref="https://wiki.duraspace.org/display/DSPACE/Helper+SQL+functions+for+DSpace+5">DSpace 5 SQL helper function</a> to resolve ID to handle):</li>
</ul>
<pre><code>for id in $(psql -U postgres -d dspacetest -h localhost -c "SELECT resource_id FROM metadatavalue WHERE resource_type_id=2 AND metadata_field_id=228 AND text_value LIKE '%SWAZILAND%'" | grep -oE '[0-9]{3,}'); do
<li>Then I couldn’t figure out a clever way to join all the CSVs, so I just grepped them to find the IDs with dates from 2018 and 2019 and there are apparently only three:</li>
<li>And looking at those items more closely, only one of them has an <em>issue date</em> of after 2018-04, so I will only update that one (as the countrie’s name only changed in 2018-04)</li>
<li>Interestingly, I see a pattern of these errors increasing, with single and double digit numbers over the past month, <del>but spikes of over 1,000 today</del>, yesterday, and on 2019-03-08, which was exactly the first time we saw this blank page error recently</li>
<li>I didn’t see anything interesting in the PostgreSQL logs, though this stack trace from the Tomcat logs (in the systemd journal) from earlier today <em>might</em> be related?</li>
</ul>
<pre><code>SEVERE: Servlet.service() for servlet [spring] in context with path [] threw exception [org.springframework.web.util.NestedServletException: Request processing failed; nested exception is java.util.EmptyStackException] with root cause
java.util.EmptyStackException
at java.util.Stack.peek(Stack.java:102)
at java.util.Stack.pop(Stack.java:84)
at org.apache.cocoon.callstack.CallStack.leave(CallStack.java:54)
at org.apache.cocoon.servletservice.CallStackHelper.leaveServlet(CallStackHelper.java:85)
at org.apache.cocoon.servletservice.ServletServiceContext$PathDispatcher.forward(ServletServiceContext.java:484)
at org.apache.cocoon.servletservice.ServletServiceContext$PathDispatcher.forward(ServletServiceContext.java:443)
at org.apache.cocoon.servletservice.spring.ServletFactoryBean$ServiceInterceptor.invoke(ServletFactoryBean.java:264)
at org.springframework.aop.framework.ReflectiveMethodInvocation.proceed(ReflectiveMethodInvocation.java:172)
at org.springframework.aop.framework.JdkDynamicAopProxy.invoke(JdkDynamicAopProxy.java:202)
at com.sun.proxy.$Proxy90.service(Unknown Source)
at org.dspace.springmvc.CocoonView.render(CocoonView.java:113)
at org.springframework.web.servlet.DispatcherServlet.render(DispatcherServlet.java:1180)
at org.springframework.web.servlet.DispatcherServlet.doDispatch(DispatcherServlet.java:950)
at org.springframework.web.servlet.DispatcherServlet.doService(DispatcherServlet.java:852)
at org.springframework.web.servlet.FrameworkServlet.processRequest(FrameworkServlet.java:882)
at org.springframework.web.servlet.FrameworkServlet.doGet(FrameworkServlet.java:778)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:624)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:731)
at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:303)
at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:208)
at org.dspace.rdf.negotiation.NegotiationFilter.doFilter(NegotiationFilter.java:59)
at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:241)
at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:208)
at org.dspace.utils.servlet.DSpaceWebappServletFilter.doFilter(DSpaceWebappServletFilter.java:78)
at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:241)
at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:208)
at org.apache.catalina.core.StandardWrapperValve.invoke(StandardWrapperValve.java:219)
at org.apache.catalina.core.StandardContextValve.invoke(StandardContextValve.java:110)
at org.apache.catalina.authenticator.AuthenticatorBase.invoke(AuthenticatorBase.java:494)
at org.apache.catalina.core.StandardHostValve.invoke(StandardHostValve.java:169)
at org.apache.catalina.valves.ErrorReportValve.invoke(ErrorReportValve.java:104)
at org.apache.catalina.valves.CrawlerSessionManagerValve.invoke(CrawlerSessionManagerValve.java:234)
at org.apache.catalina.valves.AccessLogValve.invoke(AccessLogValve.java:1025)
at org.apache.catalina.core.StandardEngineValve.invoke(StandardEngineValve.java:116)
at org.apache.catalina.connector.CoyoteAdapter.service(CoyoteAdapter.java:445)
at org.apache.coyote.http11.AbstractHttp11Processor.process(AbstractHttp11Processor.java:1137)
at org.apache.coyote.AbstractProtocol$AbstractConnectionHandler.process(AbstractProtocol.java:637)
at org.apache.tomcat.util.net.JIoEndpoint$SocketProcessor.run(JIoEndpoint.java:317)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at org.apache.tomcat.util.threads.TaskThread$WrappingRunnable.run(TaskThread.java:61)
at java.lang.Thread.run(Thread.java:748)
</code></pre>
<ul>
<li>For now I will just restart Tomcat…</li>
<li>Last week Felix from Earlham said that they finished testing on DSpace Test (linode19) so I made backups of some things there and re-deployed the system on Ubuntu 18.04
<li>During re-deployment I hit a few issues with the <ahref="https://github.com/ilri/rmg-ansible-public">Ansible playbooks</a> and made some minor improvements</li>
<li>There seems to be an <ahref="https://bugs.launchpad.net/ubuntu/+source/nodejs/+bug/1794589">issue with nodejs’s dependencies now</a>, which causes npm to get uninstalled when installing the certbot dependencies (due to a conflict in libssl dependencies)</li>
<li>I re-worked the playbooks to use Node.js from the upstream official repository for now</li>
<li>Re-sync DSpace Test with a fresh database snapshot and assetstore from CGSpace
<ul>
<li>After restarting Tomcat, Solr was giving the “Error opening new searcher” error for all cores</li>
<li>I stopped Tomcat, added <code>ulimit -v unlimited</code> to the <code>catalina.sh</code> script and deleted all old locks in the DSpace <code>solr</code> directory and then DSpace started up normally</li>
<li>I’m still not exactly sure why I see this error and if the <code>ulimit</code> trick actually helps, as the <code>tomcat7.service</code> has <code>LimitAS=infinity</code> anyways (and from checking the PID’s limits file in <code>/proc</code> it seems to be applied)</li>
<li>Then I noticed that the item displays were blank… so I checked the database info and saw there were some unfinished migrations</li>
<li>I’m not entirely sure if it’s related, but I tried to delete the old migrations and then force running the ignored ones like when we upgraded to <ahref="/cgspace-notes/2018-06/">DSpace 5.8 in 2018-06</a> and then after restarting Tomcat I could see the item displays again</li>
<li>I copied the 2019 Solr statistics core from CGSpace to DSpace Test and it works (and is only 5.5GB currently), so now we have some useful stats on DSpace Test for the CUA module and the dspace-statistics-api</li>
<li>I noticed that the regular expression for validating lines from input files in my <code>agrovoc-lookup.py</code> script was skipping characters with accents, etc, so I changed it to use the <code>\w</code> character class for words instead of trying to match <code>[A-Z]</code> etc…
<ul>
<li>We have a Spanish and French subjects so this is very important</li>
<li>Also there were some subjects with apostrophes, dashes, and periods… these are probably invalid AGROVOC subject terms, but we should save them to the rejects file instead of skipping them nevertheless</li>
</ul></li>
<li>Dump top 1500 subjects from CGSpace to try one more time to generate a list of invalid terms using my <code>agrovoc-lookup.py</code> script:</li>
</ul>
<pre><code>dspace=# \COPY (SELECT DISTINCT text_value, count(*) FROM metadatavalue WHERE metadata_field_id = 57 AND resource_type_id = 2 GROUP BY text_value ORDER BY count DESC LIMIT 1500) to /tmp/2019-03-18-top-1500-subject.csv WITH CSV HEADER;
<li>Create and merge a pull request to update the controlled vocabulary for AGROVOC terms (<ahref="https://github.com/ilri/DSpace/pull/416">#416</a>)</li>
<li>We are getting the blank page issue on CGSpace again today and I see a <del>large number</del> of the “SQL QueryTable Error” in the DSpace log again (last time was 2019-03-15):</li>
<li>Though WTF, this grep seems to be giving weird inaccurate results actually, and the real number of errors is much lower if I exclude the “binary file matches” result with <code>-I</code>:</li>
<li>It seems to be something with grep doing binary matching on some log files for some reason, so I guess I need to always use <code>-I</code> to say binary files don’t match</li>
<li>Anyways, the full error in DSpace’s log is:</li>
<li>I looked in the PostgreSQL logs, but all I see are a bunch of these errors going back two months to January:</li>
</ul>
<pre><code>2019-01-13 06:25:13.062 CET [9157] postgres@template1 ERROR: column "waiting" does not exist at character 217
</code></pre>
<ul>
<li>This is unrelated and apparently due to <ahref="https://github.com/munin-monitoring/munin/issues/746">Munin checking a column that was changed in PostgreSQL 9.6</a></li>
<li>I suspect that this issue with the blank pages might not be PostgreSQL after all, perhaps it’s a Cocoon thing?</li>
<li>Looking in the cocoon logs I see a large number of warnings about “Can not load requested doc” around 11AM and 12PM:</li>
<li>I found a handful of AGROVOC subjects that use a non-breaking space (0x00a0) instead of a regular space, which makes for a pretty confusing debugging…</li>
<li>I will replace these in the database immediately to save myself the headache later:</li>
</ul>
<pre><code>dspace=# SELECT count(text_value) FROM metadatavalue WHERE resource_type_id=2 AND metadata_field_id = 57 AND text_value ~ '.+\u00a0.+';
count
-------
84
(1 row)
</code></pre>
<ul>
<li>Perhaps my <code>agrovoc-lookup.py</code> script could notify if it finds these because they potentially give false negatives</li>
<li>CGSpace (linode18) is having problems with Solr again, I’m seeing “Error opening new searcher” in the Solr logs and there are no stats for previous years</li>
<li>Apparently the Solr statistics shards didn’t load properly when we restarted Tomcat <em>yesterday</em>:</li>
<li>For reference, I don’t see the <code>ulimit -v unlimited</code> in the <code>catalina.sh</code> script, though the <code>tomcat7</code> systemd service has <code>LimitAS=infinity</code></li>
<li>The limits of the current Tomcat java process are:</li>
</ul>
<pre><code># cat /proc/27182/limits
Limit Soft Limit Hard Limit Units
Max cpu time unlimited unlimited seconds
Max file size unlimited unlimited bytes
Max data size unlimited unlimited bytes
Max stack size 8388608 unlimited bytes
Max core file size 0 unlimited bytes
Max resident set unlimited unlimited bytes
Max processes 128589 128589 processes
Max open files 16384 16384 files
Max locked memory 65536 65536 bytes
Max address space unlimited unlimited bytes
Max file locks unlimited unlimited locks
Max pending signals 128589 128589 signals
Max msgqueue size 819200 819200 bytes
Max nice priority 0 0
Max realtime priority 0 0
Max realtime timeout unlimited unlimited us
</code></pre>
<ul>
<li>I will try to add <code>ulimit -v unlimited</code> to the Catalina startup script and check the output of the limits to see if it’s different in practice, as some wisdom on Stack Overflow says this solves the Solr core issues and I’ve superstitiously tried it various times in the past
<ul>
<li>The result is the same before and after, so <em>adding the ulimit directly is unneccessary</em> (whether or not unlimited address space is useful or not is another question)</li>
</ul></li>
<li>For now I will just stop Tomcat, delete Solr locks, then start Tomcat again:</li>
<li>After restarting I confirmed that all Solr statistics cores were loaded successfully…</li>
<li>Another avenue might be to look at point releases in Solr 4.10.x, as we’re running 4.10.2 and they released 4.10.3 and 4.10.4 back in 2014 or 2015
<ul>
<li>I see several issues regarding locks and IndexWriter that were fixed in Solr and Lucene 4.10.3 and 4.10.4…</li>
<li>It’s been two days since we had the blank page issue on CGSpace, and looking in the Cocoon logs I see very low numbers of the errors that we were seeing the last time the issue occurred:</li>
<li>To investigate the Solr lock issue I added a <code>find</code> command to the Tomcat 7 service with <code>ExecStartPre</code> and <code>ExecStopPost</code> and noticed that the lock files are always there…
<ul>
<li>Perhaps the lock files are less of an issue than I thought?</li>
<li>I will share my thoughts with the dspace-tech community</li>
<li>In other news, I notice that that systemd always thinks that Tomcat has failed when it stops because the JVM exits with code 143, which is apparently normal when processes gracefully receive a SIGTERM (128 + 15 == 143)
<li>Looking at the Cocoon log around that time I see the full error is:</li>
</ul>
<pre><code>2019-03-22 21:21:34,378 WARN org.apache.cocoon.components.xslt.TraxErrorListener - Can not load requested doc: unknown protocol: cocoon at jndi:/localhost/themes/CIAT/xsl/../../0_CGIAR/xsl//aspect/artifactbrowser/common.xsl:141:90
</code></pre>
<ul>
<li>A few milliseconds before that time I see this in the DSpace log:</li>
org.postgresql.util.PSQLException: This statement has been closed.
at org.postgresql.jdbc.PgStatement.checkClosed(PgStatement.java:694)
at org.postgresql.jdbc.PgStatement.getMaxRows(PgStatement.java:501)
at org.postgresql.jdbc.PgStatement.createResultSet(PgStatement.java:153)
at org.postgresql.jdbc.PgStatement$StatementResultHandler.handleResultRows(PgStatement.java:204)
</code></pre>
<ul>
<li>I restarted Tomcat and now the item displays are working again for now</li>
<li>I am wondering if this is an issue with removing abandoned connections in Tomcat’s JDBC pooling?
<ul>
<li>It’s hard to tell because we have <code>logAbanded</code> enabled, but I don’t see anything in the <code>tomcat7</code> service logs in the systemd journal</li>
<li><p>I sent another mail to the dspace-tech mailing list with my observations</p></li>
<li><p>I spent some time trying to test and debug the Tomcat connection pool’s settings, but for some reason our logs are either messed up or no connections are actually getting abandoned</p></li>
<li><p>I compiled this <ahref="https://github.com/gnosly/TomcatJdbcConnectionTest">TomcatJdbcConnectionTest</a> and created a bunch of database connections and waited a few minutes but they never got abandoned until I created over <code>maxActive</code> (75), after which almost all were purged at once</p>
<ul>
<li>So perhaps our settings are not working right, but at least I know the logging works now…</li>
<li>I did some more tests with the <ahref="https://github.com/gnosly/TomcatJdbcConnectionTest">TomcatJdbcConnectionTest</a> thing and while monitoring the number of active connections in jconsole and after adjusting the limits quite low I eventually saw some connections get abandoned</li>
<li>I forgot that to connect to a remote JMX session with jconsole you need to use a dynamic SSH SOCKS proxy (as I originally <ahref="/cgspace-notes/2017-11/">discovered in 2017-11</a>:</li>
<li>I need to remember to check the active connections next time we have issues with blank item pages on CGSpace</li>
<li>In other news, I’ve been running G1GC on DSpace Test (linode19) since 2018-11-08 without realizing it, which is probably a good thing</li>
<li>I deployed the latest <code>5_x-prod</code> branch on CGSpace (linode18) and added more validation to the JDBC pool in our Tomcat config
<ul>
<li>This includes the new <code>testWhileIdle</code> and <code>testOnConnect</code> pool settings as well as the two new JDBC interceptors: <code>StatementFinalizer</code> and <code>ConnectionState</code> that should hopefully make sure our connections in the pool are valid</li>
