What would you invest in learning, Openshift, Splunk , Infoblox?

I've doing K8S tools for the last 3 years and I am not sure which of those 3 ( openshift, splunk or infoblox) would add more value to my Skillset. At this point I don't want invest more effort in learning or mastering K8S and just want to jump into something else.

Openshift . Pros - easier to learn given my k8s exp and the $ rate is above devops rates . Cons - I don't see many jobs requiring it.

Splunk . Pros - very valuable $$$$ niche skill . Cons - Takes a lot in time/$$$$ wise to learn it .

Infoblox - I had some experience in the past , and I've seen a couple of roles in the market paying above market rate for this skillset .

What do you think ?

https://redd.it/g1c91b
@r_devops

Infrastructure Automation with Terraform and Atlantis

If anyone out there is using Jenkins to automate Terraform , Atlantis is worth taking a look. I have been using Atlantis for quite sometime now and love it. Actively using to spin up production infrastructure. Here is my medium blog if anyone is interested
https://tech.boxed.com/infrastructure-automation-with-terraform-0-12-and-atlantis-1a5725ee247e

https://redd.it/g1ftgh
@r_devops

WTF is wrong with full remote positions?

I eventually found a local office job that is currently 100% remote due to COVID, I'll go back to office after all this and will be fine.

But I'd like to discuss weird remote job market I had recent experience with.

About half a year ago I decided that it's damn cool to have fully remote position, primarily because it's possible to live in a cheap place, commute to next room and work in pajamas. Then I started my search, and it was pretty bad.

First - amazingly few positions are posted. There are more offline positions here, in Greater Detroit (that is not Bay Area guys), than remote positions in whole US.

Second - terrible conversion rate. I was getting around 1 tech interview for every 5-10 applications for traditional jobs. For online it was like 1 tech interview for 30-50 applications. It's completely crazy.

I have 15+ years of experience in system admin, both win and linux, last 4 is in DevOps/Cloud - theoretically my skills are in demand, that's not the problem I guess.

What's your experience?

https://redd.it/g1i7lo
@r_devops

Reusing Serverless Environments in your CI Pipeline to Reduce Serverless Integration Test Runtime from 40 minutes to 7

Recently my team managed to reduce the time taken to run Integration Tests on our Serverless microservice via CI from 40 minutes to 7 minutes!

We used another Serverless microservice to manage the re use of Serverless envrionments and CircleCI’s parallelisation tools. Check out my article if you’re interested!

https://medium.com/serverless-transformation/reducing-your-serverless-integration-test-runtime-on-ci-from-40-minutes-to-7-a8982a47b655

https://redd.it/g19fdg
@r_devops

CICD Pipeline for multiple services on AWS ECS with ECS ComposeX

[x-post](https://www.reddit.com/r/aws/comments/g0kezz/cicd_pipeline_for_multiple_services_on_aws_ecs/) from r/aws.

My first blog post of what I hope to be a regular series of articles around DevOps work. Feel free to submit any ideas on the [GH repo for the blog](https://github.com/lambda-my-aws/blog.ecs-composex.lambda-my-aws.io) and report any issues with [ECS ComposeX](https://github.com/lambda-my-aws/ecs_composex) or submit PRs.

Thanks!

https://redd.it/g198h5
@r_devops

How do you become certified in DevOps?

I am seeing positions that include text like the following: "A solid understanding of tools such as Chef, Python, PowerShell, Docker, Puppet, AWS CloudFormation, AWS, Ansible, and Kubernetes."

Let's say you learn some (or all) of these technologies at home, taking advantage of this time during pandemic.

How would you prove to prospective employers that you are capable with these technologies? Would you list on your resume the Udemy courses taken or books read, etc?

https://redd.it/g1brbc
@r_devops

upgrading multiple repo's/Deploying multiple projects with a multi-repo Design?

Hey everyone,

currently just testing this because of curiosity.. switched from a single repository where all deployment files for each container/vm reside to a "per service" design. Reason for this is that i want to improve the automated deployment of new services, and i think that switching from a monorepo to multiple ones could help here.

Every repo now looks kinda like this (example from a webcrawler project):

├── crawler <- sourcecode
│   └── crawler.py
└── deployment <- deployment files
├── deploy-int.yml
├── deploy-prod.yml
├── inventory <- ansible inventory directory container variables for each host
│   ├── host_vars
│   │   ├── crawler.local
│   │   │   └── vars.yml
│   │   └── database.local
│   │   └── vars.yml
│   └── inventory <- inventory file
└── roles <- ansible roles, included as a submodule from another repo

I have to say that i like this design. I am creating new repos for services with cookiecutter, this made the deployment process a lot easier. Also, since every service has its own repo, its easy to find out how a certain service has been configured and deployed. Inside the deployment folder are all the files which are needed for ansible to spin up the needed containers/vms, and depending on which role i add to the deploy-xxx.yml files they are also provisioned like needed.

But i have one problem with this setup:
There are situations where i feel like a single repo with every service repo could be really handy.

For example, how would i deploy a whole "site" with this design? Lets say everything burns down and a new host system is already set up but no containers/vms have been deployed yet. What i want is to push a single big red button which says "DEPLOY IT ALL!" and every service gets deployed. With a monorepo, this is an easy task because all the inventory files are in the same directory. But how do i get those when each service has its own repo? Also, how do i keep them in sync?
Not saying that a whole infrastructure burns down regularly, but i want a solution for this just in case..
I think what i need is a master repo which includes all the service repos, but not sure how i should keep it up to date. Maybe a webhook inside the deployment process of new services could do the trick?

Has someone here been in a similar situation? Or can someone think of an better solution to this? Drone is being used as a ci/cd system, maybe a site deployment can be triggered from there somehow?

Any help and hints appreciated, thanks in advance! :)
Also, sorry for any wording or grammatical errors.

https://redd.it/g1572m
@r_devops

Export amazon image into IBM cloud

Hello,

I've a VM running in AWS and want to move it to IBM Cloud. Is this supported yet? I would like to use the Amazon AMI to bring a new VM in IBM Cloud. TIA

https://redd.it/g188j1
@r_devops

Grafana - aggregating traffic graphs

Hi,

I've got some Juniper routers doing streaming telemetry into Grafana (using OpenNTI), and have currently got traffic graphs in bps per interface.

I'd like to create a single aggregate traffic graph, showing the combined inbound/outbound values of say three interfaces.

Does anybody know how to do this?

https://redd.it/g149wf
@r_devops

Telegraf/Prometheus/Grafana snmp monitoring

Hi,

I'm trying to do some SNMP monitoring for some a Cisco ASA device, I've figured out the SNMP part with Telegraf, but the next part is putting this into Prometheus, and then Grafana.

The SNMP data is being polled using an `snmp.table`, as there are multiple values in a list format.

Here is my telegraf.conf:

[agent]
hostname = "myhostname"

[global_tags]
[[inputs.snmp]]
agents = ["mydevice:161"]
version = 2
community = "mycommunitystring"
name = "snmp"
name_suffix = "_parsed"
[[inputs.snmp.field]]
name = "ike_tunnels"
oid = "iso.3.6.1.4.1.9.9.171.1.2.1.1.0"
[[inputs.snmp.field]]
name = "ipsec_tunnels"
oid = "iso.3.6.1.4.1.9.9.171.1.3.1.1.0"
[[inputs.snmp.field]]
name = "hostname"
oid = "iso.3.6.1.2.1.1.5.0"
is_tag = true
[[inputs.snmp.table]]
inherit_tags = ["hostname"]
name = "snmp"
[[inputs.snmp.table.field]]
name = "vpn_peer_ip"
oid = "iso.3.6.1.4.1.9.9.171.1.2.2.1.7"
conversion = "ipaddr"

[[outputs.prometheus_client]]
listen = myprometheusserver:9222"
metric_version = 2

This will output to this:

2020-04-14T08:43:23Z I! Starting Telegraf 1.14.0
> snmp_parsed,agent_host=mydevicehostname,host=myhostname,hostname=ciscodevicename ike_tunnels=43i,ipsec_tunnels=49i 1586853803000000000
> snmp_parsed,agent_host=mydevicehostname,host=myhostname,hostname=ciscodevicename vpn_peer_ip="1.1.1.1" 1586853803000000000
> snmp_parsed,agent_host=mydevicehostname,host=myhostname,hostname=ciscodevicename vpn_peer_ip="2.2.2.2" 1586853803000000000
> snmp_parsed,agent_host=mydevicehostname,host=myhostname,hostname=ciscodevicename vpn_peer_ip="3.3.3.3" 1586853803000000000

Prometheus will pick up the first data which includes `ike_tunnels` and `ipsec_tunnels` but it will skip the table data, which is a list of `vpn_peer_ip`, and I think that is because it is in the list format. See screenshot:

[Prometheus output screenshot](https://pasteboard.co/J3KhaqQ.png)

Meaning in grafana, I can query `snmp_parsed_ike_tunnels` and also `snmp_parsed_ipssec_tunnels` so why is it that I cannot capture the snmp table, assuming this would be `snmp_parsed_vpn_peer_ip`?

Now, I am not sure if my telegraf.conf is invalid/incorrect for what I want it to do, or does Prometheus not support this? I've spent quite a few days on this already, and I feel like I've run out of ideas, so looking for some help/direction.

Thanks.

https://redd.it/g123fn
@r_devops

Three-tier architecture: Log management

Hello,

I'm trying to understand log management tools. Can someone just explain how these stacks are built? Frontend is easy, but what about the database, the shiptool and the processing? What does Graylog do? What is the difference between EFK and ELK (okay, Logstash and Fluentd, but Logstash is again an independent log management solution).

I am happy about your answers :)

https://redd.it/g10mew
@r_devops

Pipeline and architecture drawing tools

What tool do you use to draw and graph your automation pipelines and architectures?

https://redd.it/g1y4gb
@r_devops

[Question] Running SaaS agent on K8s cluster

I'm working on a SaaS that helps K8s users with observability and management. It also provides correlated view across clusters. For that purpose there is a need for some agents to be running within the cluster and reporting metrics to the SaaS backend. The rest of the things is accessible through web site once the data reported to the SaaS service. The product uses read-only access to non sensitive cluster APIs.

I want to remove barriers for entry and simplify the use as much as possible. If you're a Kubernetes I'd appreciate your feedback on how would you like to run such SaaS service (assuming you know what it is and need it).

Have few options in mind. Please feel free to suggest anything beyond those:

1. SaaS web site uses a wizard and asks if you're a public cloud user. If in AWS/GCP/Azure will show instructions to create service account. The SaaS will then connect to client's account, get list of managed K8s cluster. User selects a cluster and the SaaS deploys necessary agents on the cluster using the namespace of the choice. For self managed k8s clusters, asks for ip address/credentials and does the same thing. No direct management, configuration and deployment is needed.
2. User deploys a helm chart which contains an operator that handles agent deployment and version updates. It has write access to its own namespace only. Deploy the operator once, the rest is automated for life.
3. User deploys, upgrades and configures agent manually. Can be involved.

https://redd.it/g1yfvs
@r_devops

Jenkins and running Projects/Builds as another user?

Hi everyone,

Quick question here. So, I built out our team's Jenkins instance about a year ago. We have an Ent Git that I use to push PowerShell scripts to and have Jenkins jobs that pull and run the scripts in Ent Git's remote repo. Simple enough and everything works great! But...

When I setup Jenkins, I have the service currently running with the "Log On" as my AD \_adm account that has Account Operator. I recently created a service account (srv\_jenkins) to have as the account in the Jenkins service "Log On". Worked just fine. I could log in to Jenkins web UI just fine. But wait... I go to run a job, specifically this job, like many of the ones, query AD/DNS/DHCP/other directory services or modify AD Objects. Well the job failed. I'm assuming that because the jobs run as the user account in the "Log On" tab for the Jenkins service? Is this correct?

If so, what is the work around, if I don't want my AD \_adm account to be running the Jenkins service, but need the jobs that would require Account Operator/Domain Admin to run the PowerShell scripts to run? Thanks!

https://redd.it/g1xs3b
@r_devops

One Size Fits None

As somebody who has been involved with many projects, included some "One Size Fits All" pipelines. I thought I'd document some problems I've ran into, as well as some things that help this transition. Hopefully this helps at least one of you in the same boat.

https://medium.com/devops-dudes/one-size-fits-none-22cf9a1725ba

https://redd.it/g1w1bo
@r_devops

Workflow for Kubernetes DevOps

[Workflow for Kubernetes DevOps that can evolve.](https://medium.com/@imarunrk/workflow-for-kubernetes-devops-15f0dbb560ff)

The developers, application/cluster operators, architects, and security team wants to contribute to the Kubernetes YAML continuously to keep the Infrastructure matching to the evolving organization strategy and policy, which demands a workflow. [Read More](https://medium.com/@imarunrk/workflow-for-kubernetes-devops-15f0dbb560ff)

https://redd.it/g229e4
@r_devops

What's the difference between Configuration Management tool and a Provisioning tool?

I hear this frequently, but can't find a decent definition. For instance, "Puppet is more of a configuration management tool, whereas Ansible is more of a provisioning, configuration and deployment tool. "

Redhat says: "Provisioning is not the same thing as configuration", and then in the next paragraph: " Server provisioning includes setting up the physical hardware in a datacenter, installing and configuring software, ..."

https://redd.it/g20rv3
@r_devops

Cloud agnostic vs using specific

Do you strive to build and deploy your apps as cloud agnostic? Or do you see value in using a specific cloud provider and leverage benefits of their managed services/unique offerings or neat integrations (*e.g. using AWS cloudformation, SAM, lambda, fargate dynamoDB etc.*)?

by cloud agnostic I mean doing everything across your software stack to be able to move the apps to other cloud provider e.g. - using abstractions in your software or configurations not to directly use specific cloud service, using terraform for IaC, serverless framework, kubernetes for orchestration, anthos? etc.

**what informs your decision to go cloud agnostic or cloud specific?**

**which technologies/tools/ frameworks promote cloud agnostic/portable cloud native engineering?**

&#x200B;

Personally, I see no clear answer, but rather take it as a project specific decision

Cost of cloud agnostic is e.g. increased operational cost, infrastructure cost, loss of speed by not using cloud provider specific innovation, reduced engineering speed for a need of maintaining layers in between your software and cloud providers

A motivation for cloud agnostic can be though: need to move workcloud across clouds, avoid vendor lock-in / compliance to enterprise strategy , legal / customer requirements

Happy to hear your thoughts!

https://redd.it/g21kga
@r_devops

Why does sorting by New not work properly in this sub?

When I sort by New in this sub only, I get a reverse sort starting about 14 hours ago. Scrolling down I see newer posts up until about 2 hours ago. After that, it starts again at about 11 hours, down to 2 or 3. This pattern repeats a few times before 1d old posts appear, and then a few 2d old posts, then some more 1d old posts.

Most other subs I'm on (haven't checked them all) behave as one would expect.

https://redd.it/g21wr8
@r_devops

Rsyslog and Journald - Log processing in Linux

Hey there,

I do not really understand the forwarding from Journald to Rsyslog.

Basically I understood it in the way that the 'pipeline' is built up as follows:

>Kernel logs through printk() → /proc/kmesg ← rsyslog → writes to log file according to rules in rsyslog.conf
Userspace logs → /dev/log ← rsyslog → writes to log file according to rules in rsyslog.conf


This results in the various log files, such as /var/log/syslog etc.

As in the rsyslog.conf snippet, messages from the facility 'user' are also written to syslog and are therefore stored twice, ***right?***

# # First some standard log files.
Log by facility.
#
auth,authpriv.* /var/log/auth.log
*.*;auth,authpriv.none -/var/log/syslog
#cron.* /var/log/cron.log
daemon.* -/var/log/daemon.log
kern.* -/var/log/kern.log
lpr.* -/var/log/lpr.log
mail.* -/var/log/mail.log
user.* -/var/log/user.log

So far so clear, if I've got it right.

Journald forwards the logs to syslog (standard in Debian according to [https://manpages.debian.org/testing/manpages-de/journald.conf.5.de.html](https://manpages.debian.org/testing/manpages-de/journald.conf.5.de.html)).

If both rsyslog and journald read from /dev/log (or the link and socket provided for syslog) and journald on Debian has ForwardToSyslog enabled by default, ***why don't all messages duplicate?***

I thank you in advance and hope that you can help me.

https://redd.it/g20mte
@r_devops

A CloudNative conversation

I’ve been working on an always evolving area of systems and infrastructure engineering over the last decade, which started for me as Automating Everything possible to doing DevOps to Everything as Code and now doing CloudNative stuff. I’ve had many conversations with myself and others on this path and here are those voices in form of many frequently asked practical questions and their answers here [https://medium.com/@ankur.floss/nativity-in-the-cloud-f14d9b197306](https://medium.com/@ankur.floss/nativity-in-the-cloud-f14d9b197306).

https://redd.it/g235as
@r_devops