Add notes for 2018-09-21

content/posts/2018-09.md

@@ -368,5 +368,30 @@ dspace=# select item_id from item where in_archive is True and withdrawn is Fals
 ## 2018-09-20
 
 - Contact Atmire to ask how we can buy more credits for future development
+- I researched the Solr `filterCache` size and I found out that the formula for calculating the potential memory use of **each entry** in the cache is:
+
+```
+((maxDoc/8) + 128) * (size_defined_in_solrconfig.xml)
+```
+
+- Which means that, for our statistics core with *149 million* documents, each entry in our `filterCache` would use 8.9 GB!
+
+```
+((149374568/8) + 128) * 512 = 9560037888 bytes (8.9 GB)
+```
+
+- So I think we can forget about tuning this for now!
+- [Discussion on the mailing list about `filterCache` size](http://lucene.472066.n3.nabble.com/Calculating-filterCache-size-td4142526.html)
+- [Article discussing testing methodology for different `filterCache` sizes](https://docs.google.com/document/d/1vl-nmlprSULvNZKQNrqp65eLnLhG9s_ydXQtg9iML10/edit)
+- Discuss Handle links on Twitter with IWMI
+
+## 2018-09-21
+
+- I see that there was a nice optimization to the ImageMagick PDF CMYK detection in the upstream `dspace-5_x` branch: [DS-3664](https://github.com/DSpace/DSpace/pull/2204)
+- The fix will go into DSpace 5.10, and we are currently on DSpace 5.8 but I think I'll cherry-pick that fix into our `5_x-prod` branch:
+  - 4e8c7b578bdbe26ead07e36055de6896bbf02f83: ImageMagick: Only execute "identify" on first page
+- I think it would also be nice to cherry-pick the fixes for [DS-3883](https://github.com/DSpace/DSpace/pull/2020), which is related to optimizing the XMLUI item display of items with many bitstreams
+  - a0ea20bd1821720b111e2873b08e03ce2bf93307: DS-3883: Don't loop through original bitstreams if only displaying thumbnails 
+  - 8d81e825dee62c2aa9d403a505e4a4d798964e8d: DS-3883: If only including thumbnails, only load the main item thumbnail.
 
 <!-- vim: set sw=2 ts=2: -->