[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
[ee4j-build] Summary of this morning's Build Team meeting
|
Arindam -- working on converting build system to Jenkins pipeline type
CI job. No CloudBees access yet. Hopefully, finished next week. Had some
access questions, but these seem to be resolved. Local builds use Maven.
In K8S, builds are 30-45 minutes. In CloudBees, may take longer. Info.
from Romain: Maven cache held in persistent claim. Can share maven
repositry/cache. Otherwise, each build will have to recreate their local
maven cache for each build (so Cache is not useful). Local Maven proxy
"near" the build systems. Eclipse does have this. Can cache include
maven.java.net (will we decommission? No.) Eclipse can create a second
proxy for maven.java.net. Nexus open-source proxy.
GF uses docker. CTS / TCKs don't. Biggest issue is creating the Docker
image. Arindam will work with Bhat and Romain to create Docker image
builds/tests for CTS and TCKs. Will run some experiments to see how this
can be done in preliminary fashion (CTS/TCK repositories aren't public).
Maybe we can use private Oracle GH organization. Docker image issues
seem readily solvable. Then need to work on getting build pipelined.
(Jenkins image/clone repository, etc.)
Will provide update next week on progress.
Discussions regarding how to create new Jenkins instance. Romain walked
through New instance, then walk through the web-ui to create new
instance. will create new instance/work with support. GH API / token/key
maybe needed.
200 Jenkins instances on current stack. Intend to migrate all to
CloudBees. Will require some YAML changes. Worker nodes should try to
limit to 4 CPU/8GB RAM. There are no quotas, but these will be added.
Will add this to the YAML data. Can one log into worker nodes? (not
accessible directly.) No Kube-ctl is planned. Romain is concerned about
this. This will require additional hardening so that any error is caught
and an archive is created, no matter what the failure. Cluster nodes are
managed, but there are restrictions due to CloudBees.
Trying to create pipeline, but failing. then successfully created.
Templates that are misconfigured can fail in an opaque fashion if they
fail in the script. (See example log output below in chat from Denis).
Oracle (Romain and Arindam) will continue to see if/how we can preserve
these details to facilitate debugging.
Can they precreate PVC (Persistent Volume Claim) resources (we don't
have credentials with perm. to generate)? Not clear. Will use "default"
storage for now.
Use Eclipse issues when problems are encountered so that these can be
handled more quickly (don't message/e-mail directly).
Create an issue for follow-up.
Chat from session:
07:02:17 From Mikaël Barbero : https://jenkins.eclipse.org/glassfish/
07:31:59 From Romain Grecourt :
https://jenkins.eclipse.org/glassfish/blue/pipelines
07:38:23 From Denis Roy : <invalid> <invalid> 1
kubelet, osnode-vm2.eclipse.org Warning
FailedMount Unable to mount volumes for pod
"mypod-95q3t-kg5g2_cje(33b4223a-7497-11e8-91ae-525400f11ca1)": timeout
expired waiting for volumes to attach/mount for pod
"cje"/"mypod-95q3t-kg5g2". list of unattached/unmounted
volumes=[maven-repo-storage]
07:44:22 From Mikaël Barbero :
https://bugs.eclipse.org/bugs/enter_bug.cgi?product=Community&component=CI-Jenkins
