Run Memory Tests on Release Candidate

[djbrooke]

We have tests from a previous release (https://github.com/IQSS/dataverse/blob/develop/scripts/tests/ec2-memory-benchmark/README.md), we should use them on the release candidate.

---

Tasks (edited by @pdurbin):

• run analysis on Dataverse 4.20
• run analysis on Dataverse 5 release candidate
• change Glassfish to Payara in ec2-memory-benchmark-remote.sh
• make sure ec2config_memory_test.yaml still works use main.yml instead
• if necessary troubleshoot timeouts with loading sample data Using this branch: IQSS/dataverse-sample-data@788efdd
• Consider changing "Utilisation" to "Utilization" to be more American. 🇺🇸

Results (edited by @pdurbin):

Dataverse 4.20

memory-benchmark-raw-dataset.txt

memory-benchmark-raw-dataverse.txt

4.20 data at https://drive.google.com/drive/folders/1VhG8PJt-LK135pPucEkE7roKaezRq8je

Dataverse 5.0 Release Candidate

memory-benchmark-raw-dataset.txt

memory-benchmark-raw-dataverse.txt

5.0rc-78527a291 data at https://drive.google.com/drive/folders/1-F65N4yNEIyx0vz5OIfLfYJHx1zgB8h4?usp=sharing

---

Config

Sample data enabled (new branch with just the data needed). Rserve installation off because it's slow to install all the R modules.

$ diff -u main.yml main2.yml 
--- main.yml	2020-07-31 14:38:14.000000000 -0400
+++ main2.yml	2020-07-30 17:04:20.000000000 -0400
@@ -130,10 +130,10 @@
   previewers:
     enabled: true
   sampledata:
-    enabled: false
+    enabled: true
     dir: /tmp/sampledata
     repo: https://github.com/IQSS/dataverse-sample-data.git
-    branch: master
+    branch: 7089-memory-tests
     venv: /tmp/sampledata/venv
   custom_sampledata:
     enabled: false
@@ -210,7 +210,7 @@
   user: prometheus
 
 rserve:
-  install: true
+  install: false
   host: localhost
   user: rserve
   group: rserve

Troubleshooting

Frequent "broken pipe" timeouts. An example:

Beginning the load test of the dataverse page.
You can monitor the progress of the test by checking the output at http://ec2-54-90-34-181.compute-1.amazonaws.com/memory-benchmark-raw-dataverse.txt
The basic test of 100 page loads, followed by a snapshot of the
heap space and the GC status, will be repeated 200 times total
(This means that once the test completes, the output file above will contain the 200 entries total;
so looking at the timestamps in the output should give you an idea how long it's going to take to finish)


packet_write_wait: Connection to 54.90.34.181 port 22: Broken pipe
./ec2-memory-benchmark.sh: line 74: unexpected EOF while looking for matching `''
./ec2-memory-benchmark.sh: line 99: syntax error: unexpected end of file
(venv) bash-3.2$ 

It's surprising that these error was seen so often because the script has a 4 hour timeout.