- Atmire responded about the issue with duplicate data in our Solr statistics
- They noticed that some records in the statistics-2015 core haven't been migrated with the AtomicStatisticsUpdateCLI tool yet and assumed that I haven't migrated any of the records yet
- That's strange, as I checked all ten cores and 2015 is the only one with some unmigrated documents, as according to the `cua_version` field
- I started processing those (about 411,000 records):
- Udana emailed me yesterday to ask why the CGSpace usage statistics were showing "No Data"
- I noticed a message in the Solr Admin UI that one of the statistics cores failed to load, but it is up and I can query it...
- Nevertheless, I restarted Tomcat a few times to see if all cores would come up without an error message, but had no success (despite that all cores ARE up and I can query them, _sigh_)
- I think I will move all the Solr yearly statistics back into the main statistics core
- Start testing export/import of yearly Solr statistics data into the main statistics core on DSpace Test, for example:
```
$ ./run.sh -s http://localhost:8081/solr/statistics-2010 -a export -o statistics-2010.json -k uid
$ ./run.sh -s http://localhost:8081/solr/statistics -a import -o statistics-2010.json -k uid
- I deployed Tomcat 7.0.107 on DSpace Test (CGSpace is still Tomcat 7.0.104)
- I finished migrating all the statistics from the yearly shards back to the main core
## 2020-12-05
- I deleted all the yearly statistics shards and restarted Tomcat on DSpace Test (linode26)
## 2020-12-06
- Looking into the statistics on DSpace Test after I migrated them back to the main core
- All stats are working as expected... indexing time for the DSpace Statistics API is the same... and I don't even see a difference in the JVM or memory stats in Munin other than a minor jump last week when I was processing them
- I will migrate them on CGSpace too I think
- First I will start with the statistics-2010 and statistics-2015 cores because they were the ones that were failing to load recently (despite actually being available in Solr WTF)
- I will migrate all these cores and see if it makes a difference, then probably end up migrating all of them
- I removed the statistics-2010, statistics-2015, statistics-2016, and statistics-2018 cores and restarted Tomcat and _all the statistics cores came up OK and the CUA statistics are OK_!
$ psql -c 'SELECT * FROM pg_locks pl LEFT JOIN pg_stat_activity psa ON pl.pid = psa.pid;' | wc -l
1739
```
## 2020-12-08
- Atmire sent some instructions for using the DeduplicateValuesProcessor
- I modified `atmire-cua-update.xml` as they instructed, but I get a million errors like this when I run AtomicStatisticsUpdateCLI with that configuration:
```
Record uid: 64387815-d9a7-4605-8024-1c0a5c7520e0 couldn't be processed
com.atmire.statistics.util.update.atomic.ProcessingException: something went wrong while processing record uid: 64387815-d9a7-4605-8024-1c0a5c7520e0, an error occured in the com.atmire.statistics.util.update.atomic.processor.DeduplicateValuesProcessor
at com.atmire.statistics.util.update.atomic.AtomicStatisticsUpdater.applyProcessors(SourceFile:304)
at com.atmire.statistics.util.update.atomic.AtomicStatisticsUpdater.processRecords(SourceFile:176)
at com.atmire.statistics.util.update.atomic.AtomicStatisticsUpdater.performRun(SourceFile:161)
at com.atmire.statistics.util.update.atomic.AtomicStatisticsUpdater.update(SourceFile:128)
at com.atmire.statistics.util.update.atomic.AtomicStatisticsUpdateCLI.main(SourceFile:78)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at org.dspace.app.launcher.ScriptLauncher.runOneCommand(ScriptLauncher.java:229)
at org.dspace.app.launcher.ScriptLauncher.main(ScriptLauncher.java:81)
- They responded with an updated CUA (6.x-4.1.10-ilri-RC7) that has a fix for the duplicates processor _and_ a possible fix for the database locking issues (a bug in CUASolrLoggerServiceImpl that causes an infinite loop and a Tomcat timeout)
- I deployed the changes on DSpace Test and CGSpace, hopefully it will fix both issues!
- In other news, after I restarted Tomcat on CGSpace the statistics-2013 core didn't come back up properly, so I exported it and imported it into the main statistics core like I did for the others a few days ago
- Sync DSpace Test with CGSpace's Solr, PostgreSQL database, and assetstore...
## 2020-12-09
- I was running the AtomicStatisticsUpdateCLI to remove duplicates on DSpace Test but it failed near the end of the statistics core (after 20 hours or so) with a memory error:
```
Successfully finished updating Solr Storage Reports | Wed Dec 09 15:25:11 CET 2020
- The statistics-2019 core finished processing the duplicate removal so I started the statistics-2017 core
- Peter asked me to add ONE HEALTH to ILRI subjects on CGSpace
- A few items that got "lost" after approval during the database issues earlier this week seem to have gone back into their workflows
- Abenet approved them again and they got new handles, phew
- Abenet was having an issue with the date filter on AReS and it turns out that it's the same `.keyword` issue I had noticed before that causes the filter to stop working
- I fixed the filter to use the correct field name and filed a bug on OpenRXV: https://github.com/ilri/OpenRXV/issues/63
- I checked the Solr statistics on DSpace Test to see if the Atmire duplicates remover was working, but now I see a comical amount of duplicates...
- I sent feedback about this to Atmire
- I will re-sync the Solr stats from CGSpace so we can try again...
- In other news, it has been a few days since we deployed the fix for the database locking issue and things seem much better now:
- I tried to harvest a few times on OpenRXV in the last few days and every time it appends all the new records to the items index instead of overwriting it:
- I'm going to try to [clone](https://www.elastic.co/guide/en/elasticsearch/reference/master/indices-clone-index.html) the temp index to the final one...
- First, set the `openrxv-items-temp` index to block writes (read only) and then clone it to `openrxv-items-final`:
at IncomingMessage.<anonymous> (/backend/node_modules/@elastic/elasticsearch/lib/Transport.js:232:25)
at IncomingMessage.emit (events.js:326:22)
at endReadableNT (_stream_readable.js:1223:12)
at processTicksAndRejections (internal/process/task_queues.js:84:21)
```
- But I'm not sure why the frontend doesn't show any data despite there being documents in the index...
- I talked to Moayad and he reminded me that OpenRXV uses an alias to point to temp and final indexes, but the UI actually uses the `openrxv-items` index
- I cloned the `openrxv-items-final` index to `openrxv-items` index and now I see items in the explorer UI
- The PDF report was broken and I looked in the API logs and saw this:
```console
(node:94) UnhandledPromiseRejectionWarning: Error: Error: Could not find soffice binary
at ExportService.downloadFile (/backend/dist/export/services/export/export.service.js:51:19)
at processTicksAndRejections (internal/process/task_queues.js:97:5)
```
- I installed `unoconv` in the backend api container and now it works... but I wonder why this changed...
- Skype with Abenet and Peter to discuss AReS that will be shown to ILRI scientists this week
- Peter noticed that [this item](https://hdl.handle.net/10568/110133) from the [ILRI policy and research briefs](https://cgspace.cgiar.org/handle/10568/24450) collection is missing in AReS, despite it being added one month ago in CGSpace and me harvesting on AReS last night
- The item appears fine in the REST API when I check the items in that collection
- Peter also noticed that [this item](https://hdl.handle.net/10568/110447) appears twice in AReS
- The item is _not_ duplicated on CGSpace or in the REST API
- We noticed that there are 136 items in the ILRI policy and research briefs collection according to AReS, yet on CGSpace there are only 132
- This is confirmed in the REST API (using [query-json](https://github.com/davesnx/query-json)):
```
$ http --print b 'https://cgspace.cgiar.org/rest/collections/defee001-8cc8-4a6c-8ac8-21bb5adab2db?expand=all&limit=100&offset=0' | json_pp > /tmp/policy1.json
$ http --print b 'https://cgspace.cgiar.org/rest/collections/defee001-8cc8-4a6c-8ac8-21bb5adab2db?expand=all&limit=100&offset=100' | json_pp > /tmp/policy2.json
$ query-json '.items | length' /tmp/policy1.json
100
$ query-json '.items | length' /tmp/policy2.json
32
```
- I realized that the issue of missing/duplicate items in AReS might be because of this [REST API bug that causes /items to return items in non-deterministic order](https://jira.lyrasis.org/browse/DS-3849)
- I decided to cherry-pick the following two patches from DSpace 6.4 into our `6_x-prod` (6.3) branch:
- High CPU usage when calling the collection_id/items REST endpoint
- Jira: https://jira.lyrasis.org/browse/DS-4342
- c2e6719fa763e291b81b2d61da2f8c758fe38ff3
- REST API items resource returns items in non-deterministic order
- Jira: https://jira.lyrasis.org/browse/DS-3849
- 2a2ea0cb5d03e6da9355a2eff12aad667e465433
- After deploying the REST API fixes I decided to harvest from AReS again to see if the missing and duplicate items get fixed
- I made a backup of the current `openrxv-items-temp` index just in case:
- After the re-harvest last night there were 200,000 items in the `openrxv-items-temp` index again
- I cleared the core and started a re-harvest, but Peter sent me a bunch of author corrections for CGSpace so I decided to cancel it until after I apply them and re-index Discovery
- I checked the 1,534 fixes in Open Refine (had to fix a few UTF-8 errors, as always from Peter's CSVs) and then applied them using the `fix-metadata-values.py` script:
- Since I was re-indexing Discovery anyways I decided to check for any uppercase AGROVOC and lowercase them:
```console
dspace=# BEGIN;
BEGIN
dspace=# UPDATE metadatavalue SET text_value=LOWER(text_value) WHERE dspace_object_id IN (SELECT uuid FROM item) AND metadata_field_id=57 AND text_value ~ '[[:upper:]]';
UPDATE 406
dspace=# COMMIT;
COMMIT
```
- I also updated the Font Awesome icon classes for version 5 syntax:
```console
dspace=# BEGIN;
dspace=# UPDATE metadatavalue SET text_value = REGEXP_REPLACE(text_value, 'fa fa-rss','fas fa-rss', 'g') WHERE text_value LIKE '%fa fa-rss%';
UPDATE 74
dspace=# UPDATE metadatavalue SET text_value = REGEXP_REPLACE(text_value, 'fa fa-at','fas fa-at', 'g') WHERE text_value LIKE '%fa fa-at%';
- Udana sent a report that the WLE approver is experiencing the same issue Peter highlighted a few weeks ago: they are unable to save metadata edits in the workflow
- Yesterday Atmire responded about the owningComm and owningColl duplicates in Solr saying they didn't see any anymore...
- Indeed I spent a few minutes looking randomly and I didn't find any either...
- I did, however, see lots of duplicates in countryCode_search, countryCode_ngram, ip_search, ip_ngram, userAgent_search, userAgent_ngram, referrer_search, referrer_ngram fields
- I sent feedback to them
- On the database locking front we haven't had issues in over a week and the Munin graphs look normal:
- After the Discovery re-indexing finished on CGSpace I prepared to start re-harvesting AReS by making sure the `openrxv-items-temp` index was empty and that the backup index I made yesterday was still there:
Thu Dec 17 07:35:27 CET 2020 | Query:containerItem:b049326a-0e76-45a8-ac0c-d8ec043a50c6
Error while updating
java.lang.UnsupportedOperationException: Multiple update components target the same field:solr_update_time_stamp
at com.atmire.dspace.cua.CUASolrLoggerServiceImpl$5.visit(SourceFile:1155)
at com.atmire.dspace.cua.CUASolrLoggerServiceImpl.visitEachStatisticShard(SourceFile:241)
at com.atmire.dspace.cua.CUASolrLoggerServiceImpl.update(SourceFile:1140)
at com.atmire.dspace.cua.CUASolrLoggerServiceImpl.update(SourceFile:1129)
...
```
- I sent the full stack to Atmire to investigate
- I know we've had thisi "Multiple update components target the same field" error in the past with DSpace 5.x and Atmire said it was harmless, but would nevertheless be fixed in a future update
- I was trying to export the ILRI community on CGSpace so I could update one of the ILRI author's names, but it throws an error...
- I added support for indexing community views and downloads to [dspace-statistics-api](https://github.com/ilri/dspace-statistics-api)
- I still have to add the API endpoints to make the stats available
- Also, I played a little bit with Swagger via [falcon-swagger-ui](https://github.com/rdidyk/falcon-swagger-ui) and I think I can get that working for better API documentation / testing
- Atmire sent some feedback on the DeduplicateValuesProcessor
- They confirm that it should process _all_ duplicates, not just those in `owningComm` and `owningColl`
- They asked me to try it again on DSpace Test now that I've resync'd the Solr statistics cores from production
- I started processing the statistics core on DSpace Test
## 2020-12-20
- The DeduplicateValuesProcessor has been running on DSpace Test since two days ago and it almost completed its second twelve-hour run, but crashed near the end:
