Cryostat can OOM when re-uploading JFR recordings into archives

When Cryostat is deployed in the supported scenario by using the Cryostat Operator, users can bypass the POST /api/v1/recordings HTTP handler and copy their .jfr binary data directly into the Cryostat archives in the OpenShift PVC.

Use `oc project` to determine the deployment namespace, `oc get pods` to determine the Cryostat pod name, and `oc get cryostats` to determine the Cryostat container name.

oc exec -i -n $MYNAMESPACE -c $CRYOSTAT_CONTAINER_NAME $CRYOSTAT_POD_NAME – mkdir /opt/cryostat.d/recordings.d/unlabelled # this may fail if the required "unlabelled" directory already exists. If so, proceed regardless

oc cp vertx-fib-demo-6f4775cdbf-82dvl_150mb_20211006T152006Z.jfr $CRYOSTAT_POD_NAME:/opt/cryostat.d/recordings.d/unlabelled/vertx-fib-demo-6f4775cdbf-82dvl_150mb_20211006T152006Z.jfr -c $CRYOSTAT_CONTAINER_NAME

After copying the recording directly into the PVC archives location, we can verify that Cryostat is aware of the uploaded file using httpie, curl, or wget:

$ https GET https://cryostat-sample-myproject.apps-crc.testing/api/v1/recordings Authorization:"Bearer $(oc whoami -t)"
HTTP/1.1 200 OK
Set-Cookie: ac6fe4a79b8104f3ee178c2547bbced6=b010120f48b63047a41c7183900059fc; path=/; HttpOnly; Secure; SameSite=None
cache-control: private
content-encoding: gzip
content-length: 182
content-type: text/plain
set-cookie: ac6fe4a79b8104f3ee178c2547bbced6=b010120f48b63047a41c7183900059fc; path=/; HttpOnly; Secure; SameSite=None

[

]

pushd $(mktemp -d)

git clone --depth 1 https://github.com/cryostatio/cryostat

cd cryostat

CRYOSTAT_IMAGE=registry.redhat.io/cryostat-20-tech-preview/cryostat-rhel8:2.0.0-2.1634741315 sh run.sh

In another terminal, after waiting a moment for the Cryostat container to start up,

curl -kv --progress-bar -F recording=@vertx-fib-demo-6f4775cdbf-82dvl_150mb_20211006T152006Z.jfr https://0.0.0.0:8181/api/v1/recordings | cat

After curl finishes transferring the file to Cryostat, the Cryostat container should die and exit within a few seconds.

I cannot attach my 156MB .jfr file tested with since it is too large. I collected it by running a Profiling recording with size limit of 150MB on the vertx-fib-demo sample app and letting that run for several hours, then coming back later to collect the sample. I'm not sure how long it actually takes for the recording to accumulate to that size before the size limit takes effect.