Closed marcoppenheimer closed 2 months ago
Oh, there is another thought I had:
It seems that we are only testing the trusted-certificates
behaviour, and not the trusted-ca
. Do we have a test for this? If not, I'd also like to test this scenario. If we can use a lot of code already written here, and this would be reasonably easy, I would make it part of this PR, otherwise I'm also fine to have a dedicated ticket. But I'd approach this next pulse
@deusebio
Do we have a test for this?
No we don't, mostly to save headache of writing more fiddly mTLS tests for the 'easier' case of CA trusting (vs individual cert). Happy to for it to be a separate ticket, as both K8s + VM will need it 👍🏾
It seems that we are only testing the
trusted-certificates
behaviour, and not thetrusted-ca
.
We don't, but the charm is executing the exact same code on both cases, minus a check on the relation name to get certificate or CA. It's good to have that test, but I don't think is critical IMO.
Changes Made
test: add mtls int-tests
run-mtls-producer
andget-offsets
app-charm actions takemtls-nodeport
instead ofbootstrap-server
, so the mTLS test assumes/requires NodePort to be enabledfeat: add LRU cache for K8s api calls with 2m ttl
pickle
-ing this, but was fiddly so left it outget_ttl_hash()
sets a 2m ttl for the in-memory cache, in-case of long-running event executions that may require recently-changed resultskube-apiserver
has--max-requests-inflight=400
as the default, so it's unlikely that this is the issue