Open dtrdnk opened 1 year ago
This is most likely due to NiFi 1.19.0 docker base image changed to eclipse-temurin:jre-11 from openjdk:8-jre. This changed the cacerts location from /usr/local/openjdk-8/lib/security/cacerts
to /opt/java/openjdk/lib/security/cacerts
.
This is most likely due to NiFi 1.19.0 docker base image changed to eclipse-temurin:jre-11 from openjdk:8-jre. This changed the cacerts location from
/usr/local/openjdk-8/lib/security/cacerts
to/opt/java/openjdk/lib/security/cacerts
.
Maybe you are right. What we can to do with this problem?
This is most likely due to NiFi 1.19.0 docker base image changed to eclipse-temurin:jre-11 from openjdk:8-jre. This changed the cacerts location from
/usr/local/openjdk-8/lib/security/cacerts
to/opt/java/openjdk/lib/security/cacerts
.
Yes, but with cert-manager enabled the chart overrides that default by setting the nifi.security.truststore
configuration property. And it also sets the nifi.security.autoreload.enabled
and nifi.security.autoreload.interval
properties per the documentation (https://nifi.apache.org/docs/nifi-docs/html/administration-guide.html#security_configuration).
We are also interested in upgrading to NiFi 1.18 or 1.19 for other reasons, so I'll get to work replicating the problem.
Confirmed that the certificate rotation broke between 1.16.3 and 1.17.0. Suspicious of this commit: https://github.com/apache/nifi/commit/4b655ecb688685016f67fba7a9386683f8429df6
same for 1.20, see #294
@wknickless @dtrdnk #294 fixes the issue and bumps NiFi to 1.20.
I'm happy to merge the PR but I would be more at ease with more eyes on the changes.
Hi @banzo, in my PR #294, I just reatart nifi pod to make it reload new keystore, it seems the autoreload feature is broken and has not been fixed yet
@wknickless @dtrdnk #294 fixes the issue and bumps NiFi to 1.20.
I'm happy to merge the PR but I would be more at ease with more eyes on the changes.
Good news! I wiil wait merge of PR 294. And then I will merge all commits from master into my PR
Hi @banzo, in my PR #294, I just reatart nifi pod to make it reload new keystore, it seems the autoreload feature is broken and has not been fixed yet
You are right. Autoreload function is broken in new versions of Docker image, and I don't know how to fix this problem.
@dtrdnk is one of those issues describing the problem or do we need to create a new one?
@dtrdnk is one of those issues describing the problem or do we need to create a new one?
Thank you! This issue NIFI-10425 looks like our problem. I make research by the Jetty problem, and get some facts:
I think we should just wait for the NIFI-11518 task to be released.
Hi @dtrdnk, I tested locally on windows (no symlink) machine and the autoreload feature has also broken since v1.17.0. I created NIFI-11536 for tracking this issue. The feature works for 1.16.3 so it clearly is not the issue with jetty
Describe the bug Current app version 1.16.3 works fine, but if you upgrade to 1.19.0, then the automatic rotation of certificates breaks. This breaks also breaks the test-site-to-site
Version of Helm, Kubernetes and the Nifi chart: Helm : v3.10.2 Kubernetes: v1.25.3 NiFi chart: 1.1.3
What happened: NiFi server is still working with the old certificate, even if new cert exist in the store
What you expected to happen: NiFi server must use a new certificate on the fly without restart
How to reproduce it (as minimally and precisely as possible): Just upgrade image tag from 1.16.3 to 1.19.0 or 1.18.0
Anything else we need to know:
Here are some information that help troubleshooting: NiFi has autoreload function, but I don't know how to invoke it. By default autoreload set to false. Even if I set manualy autoreload to true, this is doesn't help.
values.yaml
or the changes made to the default one (after removing sensitive information)Check if a pod is in error: There is no a pod error