Advent of Sysadmin 2025(sadservers.com) |
Advent of Sysadmin 2025(sadservers.com) |
1. Get a user to stop logging in as root.
2. Get all users to stop sharing the same login and password for all servers.
3. Get a user to upgrade their app's dependencies to versions newer than 2010.
4. Get a user to use configuration management rather than scp'ing config files from their laptop to the server.
5. Get a user to bake immutable images w/configuration rather than using configuration management.
6. Get a user to switch from Jenkins to GitHub Actions.
7. Get a user to stop keeping one file with all production secrets in S3, and use a secrets vault instead.
8. Convince a user (and management) you need to buy new servers, because although "we haven't had one go down in years", every one has faulty power supply, hard drive, network card, RAM, etc, and the hardware's so old you can't find spare parts.
9. Get management to give you the authority to force users to rotate their AWS access keys which are 8 years old.
10. Get a user to stop using the aws root account's access keys for their application.
11. Get a user to build their application in a container.
12. Get a user to deploy their application without you.
Github Actions left a bad taste in my mouth after having it randomly removed authenticated workers from the pool, after their offline for ~5 days.
This was after setting up a relatively complex PR workflow (always on cheap server starts up very expensive build server with specific hardware) only to have it break randomly after a PR didn't come in for a few days. And no indication that this happens, and no workaround from GitHub.
There are better solutions for CI, GitHub 's is half baked.
That said, I have found runners to be unnecessarily difficult.
But Jenkins and its own quirks, and when I used GitLab, it used ancient docker-machine and outdated AMIs by default.
I think Buildkite has been the only one to make this easy and scalable. But it is meant for self hosted runners.
[1] https://docs.github.com/en/enterprise-cloud@latest/actions/h...
Oh, good lord why?
I've notified the authorities and social services.
Most are obvious to most people. None are obvious to everybody.
It really depends if the machine is hosting anything that you don't want some users to access. If the machine is single-purpose and any user is already able to access everything valuable from it (DB with customer data, etc) or trivially elevate to root (via sudo, docker access, etc) then it's just pointless extra typing and security theatre.
Is this really like that? Isn't there any Unix/DBA anymore? I associate DevOps to what at my time we called "operations" and "development". We had 5 teams or so:
1) Developers, who would architect and write code, 2) Operations who would deploy, monitor and address customer complaints, 3) Unix (aka SYS) administrators, who would take care of housekeeping of well, the OS (and web servers/middleware), 4) DBA who would be monitoring and optimizing Oracle/Postgres, and 5) Network admins, who would take care of Load Balancers, Routers, Switches, Firewalls (well, there were 2 security experts for that also)
So I think DevOps would be a mix of 1&2, to avoid the daily wars that would constantly happen "THEY did it wrong!"
Can somebody clear my mind, please!? It seems I was out of it for too long?!
Developers handle 1). Devops handle 2)/3)/5). Nobody does 4)
[1] https://github.com/ankane/pghero
[2] https://pgtune.leopard.in.ua/
Edit: For the record, I have worked at a few small companies as the "SysAdmin" guy who did the whole compliment of servers, OS, storage, networking, VMs, DB, perf tuning, etc.
1. Patch Microsoft exchange with only a three hour outage window 2. Train a user to use onedrive instead of emailing 50mb files and back and forth 3. Setup eight printers for six users. Deal with 9gb printer drivers. 4. Ask an exec if he would please let you add mfa to their mailbox. 5. Sit there calmly while that exec yells like a wwe wrestler about the ways he plans to ruin you in response 6. Debate the cost of a custom mouse pad for one person across three meetings 7. Deploy any standard windows app that expects everyone be an administrator without making everyone an administrator 8. Deploy an app that expects uac disabled without disabling uac 9. Debug some finance persons 9000 line excel function
Ask?! This is where the org's cyber insurance is your friend. Just have the executive get the provider's clearance on him not having MFA. I'm sure that line item will change his mind, and if not, be sure to accidently mention those exemptions to those yearly auditors.
Deploy new Server(s), patch, install Exchange, Setup DAGs, migrate everyone mailbox, swing load balancer over to new servers, uninstall Exchange from old, remove old from Active Directory, delete servers.
BTW, Upgrades now suck because Office365 uses method above so upgrade system never gets good Q&A from them.
9. Get management to give you the authority to force users to rotate their AWS access keys which are 8 years old.
You can definitely make a strong argument that the organization needs practice rotating, so I would advise reframing it as an org-survivability-planning challenge and not a key-security issue.
Damn, this one I'm guilty of. Though, I'm not real Sysadmin/DevOps, I'm just throwing something together and deploying it on a LAN-only VM for security reasons (I don't trust the type of code I would write)
A: Calculate the average age in years of all dependencies calculated by: (max(most recent version release date, date of most recent CVE on library) - used version release date). Sleep for that many seconds before the app starts.
If you don't want a user to log in as root, disable the root password (or change it to something only you know) and disable root ssh. If you want people to stop sharing the same login and password across all servers, there's several ways to do it but the most straightforward one seems like it would be to enforce the use of a hardware key (yubikey or similar) for login. If people aren't using configuration management software and are leaving machines in an inconsistent state, again there are several options but I'd look into this NixOS project: https://github.com/nix-community/impermanence + some policy of rebooting the machines regularly.
If you don't like how users are making use of AWS resources and secrets, then set up AWS permissions to force them to do so the correct way. In general if someone is using a system in a bad or insecure way, then after alerting them with some lead time, deliberately break their workflow and force them to come to you in order to make progress. If the thing you suggest is actually the correct course of action for your organization, then it will be worthwhile.
If you just do any of this list without the proper migration plan/time, someone senior in the org will complain and you will lose.
Two pints of ale please!
Feedback from candidates is that they find it a bit stressful during the actual interview but love the approach once it's completed.
The interview option also makes it trivial to just send to a candidate via Zoom chat, ask them to share their screen and "just works".
Happy to answer questions folks may have about how we use it.
Any other suggestions? I have sysadmin experience as a homelabber and at work with a small company as a "tech lead" but have not yet had the chance to do it full time in a larger company. Currently focused on back-filling knowledge gaps and adding certs to support my existing experience.
If you are looking into more of the "people" side of things, I would HIGHLY recommend Never Split the Difference by Chris Voss [0]. A big part of being a team lead and/or working at a larger firm is understanding where people are coming from and then convincing them that your solution is "win/win". The book is great at highlighting multiple different tactics to do that.
Turn the Ship Around [1] is also great at giving examples of how to "change organizations in place". If you end up at larger firms, there will be a LOT of legacy infra and processes that you may want to improve. Marquet gives excellent examples of how to change things WHILE ALSO getting buy in from the team.
imagine typing in a terminal...
you want to delete the previous word so press ctrl+w...
actually you're in a browser; the window closes...
EDIT: Ah, ok, `vi` is installed on the server _itself_, just not in the Docker containers. So I guess I'm going to have to `docker cp` them in. Can do o7
In small companies, sysadmin might be a duty of the SRE team, but they definitely diverge if you have a large on-prem deployment or work with bespoke VMs in the cloud.
I don't know of any other SaaS which gives you a VM with one click without any registration but we do it.
In any case thanks for the feedback, I've put a button on this /advent page for clarity, cheers
> Sign up for a free account (needed to keep track of your progress)
is a complete lie. Tracking a person’s progress is what cookies are for. You don’t need us to create an account for that.
What you do need users to create accounts for, is for you to track every user and their progress.
If you tell me more, I might sign up. If I have to create an account first, I'm walking away.
I would like to see and try to solve the scenarios for myself, not to get meaningless internet points. If you look at their front page, you can do that right now. So why do I have to create an account to even see these special advent scenarios?
> do you even sysadmin?
Yes.
At 5$/m I might give the paid subscription a try.
This kind of thing annoys me. This is why CTFs are great, where the goal is to get the flag string. Obviously harder to do for sysadmin, but expecting a particular configuration when I managed to make it work without doing things exactly as they wanted is no better than a poorly written exam.
It is hard to have a checker that eliminates both false positives and false negatives in general, but we always try to minimize false negatives and we failed initially here.
The 12 days of Christmas start on Christmas and end on January 5, the eve of the Feast of Epiphany.
12-day advent calendars are a fairly recent invention that mirrors the 12-days of Christmas, but has no direct correspondence to anything in any traditional Christian religious calendar (the more common 24-day format is also a modern, but less recent, invention detached from the religious calendar, that simplifies by ignoring the floating start date of advent and always starting on Dec. 1.)
We have scenarios running on k8s, both on single VMs (the ones you can see in the scenario list) and we also have a beta/PoC k8s cluster where we currently run a couple of scenarios as single pod (a docker container) or as a full system (the "kubernetes playgrounds", which is kind of hidden while we test it).
Is this what you were wondering? we do have pending to introduce podman scenarios as well
Somehow, SadServers seems to have entirely missed the concept of a "puzzle".
More accurate statement imo.
Great suggestions and thanks for taking the time to respond :)
Advent begins on the fourth Sunday before Christmas, which was Nov 30 this year. It ends on Dec 24. Therefore it is technically anywhere from 22 to 28 days long.
Advent calendars begin on Dec 1 and end on Dec 25.
It was originally (and still usually) used by those who wanted to self-host runners.
I must say, again I thought (I read it somewhere?) DevOps should take care of the constant battle between Devs and Operations (I've seen enough of that in my times) by merging 1 and 2 together. But it seems just a name change, and if anything, seems worst, as a (IMHO) critical and central component, like the DB, now has totally distributed responsibilities. I would like to know what happens when e.g. a DB crashes because a filesystem is full, "because one developer made another index, because one from devops had a complaint because X was too slow".
Either the people are extremely more professional that in my times, or it must be a shitshow to look while eating pop-corn.
In practice there is no way to relay "query fubar, fix" back, because we are much agile, very scrum: feature is done when the ticket is closed, new tickets are handled by product owners. Reality is antithesis of that double Ouroboros.
In practice developers write code, devops deploy "teh clouds" (writing yamls is the deving part) and we throw moar servers at some cloud db when performance becomes sub-par.
Then they hire DBREs because they think DBA sounds antiquated, who then enter a hellscape of knowing exactly what the root issues are (poorly-designed schemata, unperformant queries, and applications without proper backoff and graceful degradation), and being utterly unable to convince management of this (“what if we switched to $SOME_DBAAS? That would fix it, right?”).
> bunch of random Jenkins servers
Either PXE boot from an image, or k8s from an image, have a machine or pod rebooted/destroyed after one job. Update your image once a month, or have a Jenkins job to do that for you.
> Authentication, authorization, access control
Either use LDAP or Login via Github, and Matrix security plugin. Put all "Devops" group into admins, the rest into users, never touch it again.
> configuration
CASC plugin and seed for jobs, and/or Helm for just about everything else.
> env vars and secrets
Pull everything from Vault with Vault plugin.
> as a sysadmin Jenkins will suck the life out of you
I spend about 1-2 hours a week managing Jenkins itself, and the rest of the week watching the jobs or developing new ones.
And this is if you get to manage it! Often there's 5 different random Jenkins servers set up by different teams, all of which are EOL and rife with security holes, and they expect you to fix them when they break, nobody version controls their configs or backs them up (they haven't even heard of CasC and have no interest in using it), and your boss says you can't say no, and also you can't upgrade them/take them over. I've seen million-dollar products which are completely dependent on over a thousand Jenkins jobs on an out-of-date Jenkins server, so complex and intertwined it couldn't be replaced.
If it were up to me, I would replace most CI with Drone.io (or Woodpecker CI if it ever gets feature parity). Now that's a dead simple CI system.
But yeah, if all there is to do is "git clone && build && deploy" then Jenkins is an overkill and it probably wasn't warranted in the first place.
> you can already build such a system yourself quite trivially by getting an FTP account, mounting it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem. From Windows or Mac, this FTP account could be accessed through built-in software.
I do that too, until my complex logic belongs to the CI system and nowhere else.
As an example:
It's 2 am now and Jenkins needs to run some (but not all) nightly tests[0]. To figure out which, let's bring the source code and analyse the configuration file, disregarding anything that doesn't need to run at this hour. Once we have the plan for which tests to run, let's figure out what we need to build[1]. Also, let's see what is the status of the pool running the tests[2] so we can decide on a tests parallelisation strategy for this run. When we have a plan, let's build and test, keeping an eye on the triggered tests. When all these have finished, analyse the failures[3], create tickets for the failures[4] and prepare a report to be sent.
[0] other "nightly tests" run at 1, 3, 4 etc.
[1] this is mapped in the configuration file too.
[2] this is internal to Jenkins
[3] same
[4] this involves finding the "responsible person," so a lot of API calls
Fwiw, I do believe you can do this in GHA, but you may need to call their API from your workflow. In addition, their replacement for Groovy is to run an action which lets you embed Javascript/Typescript and call their SDK. It sucks, but so does Groovy! ;-)
2. migrate secret manager. Roll again
3. cloud build
4. gocd
5. jenkins
6. gitlab
7. github actions
8. bamboo
9. codepipeline
10. buildbot
11. team foundation server
12. migrate version control. Roll againIf you use the built in container registry and build artifacts, you can pass between steps.
You need a beefy master and it is your single point of failure. Untimely triggers of heavy jobs overwhelm controller? All projects are down. Jobs need to be carefully crafted to be resumable at all.
Heavy reliance on master means that even sending out webhooks on stage status changes is extremely error prone.
When your jobs require certain tools to be available you are expected to package those as part of agent deployment as Jenkins relies on host tools. In reality you end up rolling your own tool management system that every job has to call in some canonical manner.
There is no built in way to isolate environments. You can harden the system a bit with various ACLs, but in the end if you either have to trust projects or build up and maintain infrastructures for different projects isolated at host level.
In cases when time-wise significant processing happens externally, you have to block an executor.
Will test forgejo's CI first as we'll use the repo anyway, but if it ain't for me, it's going to be jenkins I assume.
- DSL is harder to get into.
- Hard to reproduce a setup unless builds are in DSL and Jenkins itself is in a fixed version container with everything stored in easily transferable bind volumes; config export/import isn't straightforward.
- Builds tend to break in a really weird way when something (even external things like Gitea) updates.
- I've had my setup broken once after updating Jenkins and not being able to update the plugins to match the newer Jenkins version.
- Reliance on system packages instead of containerized build environment out of the box.
- Heavier on resources than some of the alternatives.
- GUI is getting prettier lately for some reason.
- Great extendability via plugins.
- A known tool for many.
- Can mostly be configured via GUI, including build jobs, which helps to get around things at first (but leads into the reproducibility trap later on).
Setting up your own Gitlab instance + Runners with rootless containers is not without quirks, too.
We have our main node in a state that we can move it anywhere in a couple of minutes with almost no downtime.
I'll add another point to "Pros": Jenkins is FOSS and it costs $0 per developer per month.
Haven't spent a lot of time with it myself, but if Jenkins isn't of much appeal, Drone [1] seems to be another popular (and lightweight) alternative.