Ah nice. I’m gonna dig through their agent code for kube and see if there’s any low hanging fruit that can be knocked out to get it talking to openshift/heapster. They have a couple repos on github https://github.com/DataDog/integrations-core/blob/master/kubernetes/check.py
Alex From: David Conde <[email protected]> Date: Thursday, May 18, 2017 at 8:56 AM To: Alex Creek <[email protected]> Cc: Jay Vyas <[email protected]>, users <[email protected]> Subject: Re: Kubelet & Cadvisor Hi Alex, I was able to get past that by enabling hostDir for the service account, I also had to set privileged: true in the security context so it could access the cgroups mounts etc. The next issue I hit was it tries to access cAdvisor via a non secure port instead of via the kubelet stats api. I also had to mount kubelet certs into a volume so that they could be used to talk to kubelet over SSL. I contacted their support about the cAdvisor issue, they responded with its currently in their backlog. On 18 May 2017 at 13:16, Alex Creek <[email protected]> wrote: +1 for protips on getting datadog to monitor openshift I wasn’t able to get the datadog agent working with 1.4. I tried both of their out-of-the-box solutions for the k8s integration and no dice. Have yet to deep dive and investigate. The initial problem I ran into was their pre-baked monitoring daemonSet used hostPath volumes and I didn’t have them enabled. Once I did enable them the entire container filesystem was readonly and the datadog agent fell over when it tried to write to disk :\ I tried the other solution running the agent via docker cli and the container starts and appears to be working but no metrics are collected. Alex From: <[email protected]> on behalf of David Conde <[email protected]> Date: Thursday, May 18, 2017 at 6:24 AM To: Jay Vyas <[email protected]> Cc: users <[email protected]> Subject: Re: Kubelet & Cadvisor Thanks Jay, I contacted DataDog and they have said it is a known issue for them which they expect to have fixed in a few months. Can anyone point me at a guide for a good setup for monitoring and alerting when using Openshift? I have looked at DataDog but it looks like there is an outstanding bug in the k8s integration which prevents it working with Openshift installs. I also installed hawkular but it looks like there is no UI to configure alerts in place yet. Would anyone be willing to share their experience? Thanks, Dave On 17 May 2017 at 13:23, Jay Vyas <[email protected]> wrote: On May 17, 2017, at 7:37 AM, David Conde <[email protected]> wrote: Hi, I am trying to get the DataDog agent working with k8s integration. I'm hitting an issue around cadvisor not being available. I have read that cadvisor is available via kubelet as long as I'm using certs to access it. Does anyone know what the equivalent of the 2 URLs below are when trying to access via kubelet? - /api/v1.3/machine/ - /api/v1.3/subcontainers/ In general I don't think there is anything wrong with accessing cadvisor directly, except that I believe cadvisor isn't exposed from kubelets.... So i think, /stats/summary in the kubelet will externalize some of the cadvisor metrics you want. The thread below describes the idea behind stats as a cadvisor wrapper https://groups.google.com/forum/m/#!topic/kubernetes-sig-node/txBjT8-WvM0 Thanks, Dave _______________________________________________ users mailing list [email protected] http://lists.openshift.redhat.com/openshiftmm/listinfo/users _______________________________________________ users mailing list [email protected] http://lists.openshift.redhat.com/openshiftmm/listinfo/users
_______________________________________________ users mailing list [email protected] http://lists.openshift.redhat.com/openshiftmm/listinfo/users
