Of all the places I've worked that had the attitude "If this goes down at 3AM, we need to fix it immediately", there was only one where that was actually justifiable from a business perspective. I'm worked at plenty of places that had this attitude despite the fact that overnight traffic was minimal and nothing bad actually happened if a few clients had to wait until business hours for a fix.
I wonder if some of the preference for big-name cloud infrastructure comes from the fact that during an outage, employees can just say "AWS (or whatever) is having an outage, there's nothing we can do" vs. being expected to actually fix it
From this perspective, the ability to fix problems more quickly when self hosting could be considered an antifeature from the perspective of the employee getting woken up at 3am
That includes big and small businesses, SaaS and non-SaaS, high scale (5M+rps) to tiny scale (100s-10krps), and all sorts of different markets and user bases. Even at the companies that were not staffed or providing a user service over night, overnight outages were immediately noticed because on average, more than one external integration/backfill/migration job was running at any time. Sure, “overnight on call” at small places like that was more “reports are hardcoded to email Bob if they hit an exception, and integration customers either know Bob’s phone number or how to ask their operations contact to call Bob”, but those are still environments where off-hours uptime and fast resolution of incidents was expected.
Between me, my colleagues, and friends/peers whose stories I know, that’s an N of high dozens to low hundreds.
What am I missing?
IME the need for 24x7 for B2B apps is largely driven by global customer scope. If you have customers in North American and Asia, now you need 24x7 (and x365 because of little holiday overlap).
That being said, there are a number of B2B apps/industries where global scope is not a thing. For example, many providers who operate in the $4.9 trillion US healthcare market do not have any international users. Similarly the $1.5 trillion (revenue) US real estate market. There are states where one could operate where healthcare spending is over $100B annually. Banks. Securities markets. Lots of things do not have 24x7 business requirements.
All of those places needed their backend systems to be up 24/7. The banks ran reports and cleared funds with nightly batches—hundreds of jobs a night for even small banking networks. The healthcare companies needed to receive claims and process patient updates (e.g. your provider’s EMR is updated if you die or have an emergency visit with another provider you authorized for records sharing—and no, this is not handled by SaaS EMRs in many cases) over night so that their systems were up to date when they next opened for business. The “regular” businesses closed for the night generated reports and frequently had IT staff doing migrations, or senior staff working on something at midnight due the next day (when the head of marketing is burning the midnight oil on that presentation, you don’t want to be the person explaining that she can’t do it because the file server hosting the assets is down all the time after hours).
And again, that’s the norm I’ve heard described from nearly everyone in software/IT that I know: most businesses expect (and are willing to pay for or at least insist on) 24/7 uptime for their computer systems. That seems true across the board: for big/small/open/closed-off-hours/international/single-timezone businesses alike.
You wake up. It's not your fault. You're helpless to solve it.
Eventually, AWS has a VP of something dial in to your call to apologize. They’re unprepared and offer no new information. The get handed to a side call for executive bullshit.
AWS comes back. Your support rep only vaguely knows what’s going on. Your system serves some errors but digs out.
Then you go to sleep.
And Postgres upgrades are not transparent. So you'll have a 1 or 2 hours task, every 6 to 18 months that you have only a small amount of control over when it happens. This is ok for a lot of people, and completely unthinkable for some other people.
I was disappointed alloy doesn't support timescaledb as a metrics endpoint. Considering switching to telegraf just because I can store the metrics on Postgres.
SQLite when prototyping, Postgres for production.
If you need to power a lawnmower and all you have is a 500bhp Scania V8, you may as well just do it.
Even better, stage to a production-like environment early, and then deploy day can be as simple as a DNS record change.
I have switched to using postgres even for prototyping once I prepared some shell scripts for various setup. With hibernate (java) or knex (Javascript/NodeJS) and with unit tests (Test Driven Development approach) for code, I feel I have reduced the friction of using postgres from the beginning.
Because it's "low effort" to just fire it into sqlite and if I have to do ridiculous things to the schema as I footer around working out exactly what I want the database to do.
I don't want to use nodejs if I can possibly avoid it and you literally could not pay me to even look at Java, there isn't enough money in the world.
-Backups: the provider will push a full generic disaster-recovery backup of my database to an off-provider location at least daily, without the need for a maintenance window
-Optimization: index maintenance and storage optimization are performed automatically and transparently
-Multi-datacenter failover: my database will remain available even if part(s) of my provider are down, with a minimal data loss window (like, 30 seconds, 5 minutes, 15 minutes, depending on SLA and thus plan expenditure)
-Point-in-time backups are performed at an SLA-defined granularity and with a similar retention window, allowing me to access snapshots via a custom DSN, not affecting production access or performance in any way
-Slow-query analysis: notifying me of relevant performance bottlenecks before they bring down production
-Storage analysis: my plan allows for #GB of fast storage, #TB of slow storage: let me know when I'm forecast to run out of either in the next 3 billing cycles or so
Because, well, if anyone provides all of that for a monthly fee, the whole "self-hosting" argument goes out of the window quickly, right? And I say that as someone who absolutely adores self-hosting...
Corollary: rental/SaaS models provide that property in large part because their providers have lots of slack.
Yes, I'd say backups and analysis are table stakes for hiring it, and multi-datacenter failover is a relevant nice to have. But the reason to do it yourself is because it's literally impossible to get anything as good as you can build in somebody's else computer.
In case you want to self host but also have something that takes care of all that extra work for you
(I do use Maria at home for legacy reasons, and have used MySQL and Pg professionally for years.)
Can you give any details on that?
I switched to MariaDB back in the day for my personal projects because (so far as I could tell) it was being updated more regularly, and it was more fully open source. (I don't recall offhand at this point whether MySQL switched to a fully paid model, or just less-open.)
It’s not like Mongo gives you those properties for free either. Replication/clustering related data loss is still incredibly common precisely because mongo makes it seem like all that stuff is handled automatically at setup when in reality it requires plenty of manual tuning or extra software in order to provide the guarantees everyone thinks it does.
CloudNativePG (https://cloudnative-pg.io) is a great option if you’re using Kubernetes.
There’s also pg_auto_failover which is a Postgres extension and a bit less complex than the alternatives, but it has its drawbacks.
Patroni has been around for awhile. The database-as-a-service team where I work uses it under the hood. I used it to build database-as-a-service functionality on the infra platform team I was at prior to that.
It's basially push-button production PG.
There's at least one decent operator framework leveraging it, if that's your jam. I've been living and dying by self-hosting everything with k8s operators for about 6-7 years now.
Yet they still call it HA because there's nothing else. Even a planned shutdown of the primary to patch the OS results in downtime, as all connections are terminated. The situation is even worse for major database upgrades: stop the application, upgrade the database, deploy a new release of the app because some features are not compatible between versions, test, re-analyze the tables, reopen the database, and only then can users resume work.
Everything in SQL/RDBMS was thought for a single-node instance, not including replicas. It's not HA because there can be only one read-write instance at a time. They even claim to be more ACID than MongoDB, but the ACID properties are guaranteed only on a single node.
One exception is Oracle RAC, but PostgreSQL has nothing like that. Some forks, like YugabyteDB, provide real HA with most PostgreSQL features.
About the hype: many applications that run on PostgreSQL accept hours of downtime, planned or unplanned. Those who run larger, more critical applications on PostgreSQL are big companies with many expert DBAs who can handle the complexity of database automation. And use logical replication for upgrades. But no solution offers both low operational complexity and high availability that can be comparable to MongoDB
>> "God Send". Everything just worked. Replication was as reliable as one could imagine. It outlives several hardware incidents without manual intervention. It allowed cluster maintenance (software and hardware upgrades) without application downtime. I really dream PostgreSQL will be as reliable as MongoDB without need of external services.
https://www.postgresql.org/message-id/0e01fb4d-f8ea-4ca9-8c9...
I would expect a little bit more as a cost of the convenience, but in my experience it's generally multiple times the expense. It's wild.
This has kept me away from managed databases in all but my largest projects.
If anything that’s a feature for ease of use and compatibility.
I know there are other issues with Kubernetes but at least its transferable knowledge.
https://blog.notmyhostna.me/posts/what-i-wish-existed-for-se...
> If you're just starting out in software & want to get something working quickly with vibe coding, it's easier to treat Postgres as just another remote API that you can call from your single deployed app
> If you're a really big company and are reaching the scale where you need trained database engineers to just work on your stack, you might get economies of scale by just outsourcing that work to a cloud company that has guaranteed talent in that area. The second full freight salaries come into play, outsourcing looks a bit cheaper.
This is funny. I'd argue the exact opposite. I would self host only:
* if I were on a tight budget and trading an hour or two of my time for a cost saving of a hundred dollars or so is a good deal; or
* at a company that has reached the scale where employing engineers to manage self-hosted databases is more cost effective than outsourcing.
I have nothing against self-hosting PostgreSQL. Do whatever you prefer. But to me outsourcing this to cloud providers seems entirely reasonable for small and medium-sized businesses. According to the author's article, self hosting costs you between 30 and 120 minutes per month (after setup, and if you already know what to do). It's easy to do the math...
except now they are stuck trying to maintain and debug Postgres without having the same visibility and agency that they would if they hosted it themselves. situation isn't at all clear.
I use Google Cloud SQL for PostgreSQL and it's been rock solid. No issues; troubleshooting works fine; all extensions we need already installed; can adjust settings where needed.
in the limit I dont think we should need DBAs, but as long as we need to manage indices by hand, think more than 10 seconds about the hot queries, manage replication, tune the vacuumer, track updates, and all the other rot - then actually installing PG on a node of your choice is really the smallest of problems you face.
This leads the developers to do all kinds of workarounds and reach for more cloud services (and then integrating them and - often poorly - ensuring consistency across them) because the cloud hosted DB is not able to handle the load.
On bare-metal, you can go a very long way with just throwing everything at Postgres and calling it a day.
Every company out there is using the cloud and yet still employs infrastructure engineers to deal with its complexity. The "cloud" reducing staff costs is and was always a lie.
PaaS platforms (Heroku, Render, Railway) can legitimately be operated by your average dev and not have to hire a dedicated person; those cost even more though.
Another limitation of both the cloud and PaaS is that they are only responsible for the infrastructure/services you use; they will not touch your application at all. Can your application automatically recover from a slow/intermittent network, a DB failover (that you can't even test because your cloud providers' failover and failure modes are a black box), and so on? Otherwise you're waking up at 3am no matter what.
Whether or not you need that equivalence is an orthogonal question.
There's probably a sweet spot where that is true, but because cloud providers offer more complexity (self-inflicted problems) and use PR to encourage you to use them ("best practices" and so on) in all the cloud-hosted shops I've been in a decade of experience I've always seen multiple full-time infra people being busy with... something?
There was always something to do, whether to keep up with cloud provider changes/deprecations, implementing the latest "best practice", debugging distributed systems failures or self-inflicted problems and so on. I'm sure career/resume polishing incentives are at play here too - the employee wants the system to require their input otherwise their job is no longer needed.
Maybe in a perfect world you can indeed use cloud-hosted services to reduce/eliminate dedicated staff, but in practice I've never seen anything but solo founders actually achieve that.
It's complexity but it's also providing features. If you didn't use those cloud features, you'd be writing or gluing together and maintaining your own software to accomplish the same tasks, which takes even more staff
> Maybe in a perfect world you can indeed use cloud-hosted services to reduce/eliminate dedicated staff
So let's put it another way: either you're massively reducing/eliminating staff to achieve the same level of functionality, or you're keeping the equivalent staff but massively increasing functionality.
The point is, clouds let you deliver a lot more with a lot less people, no matter which way you cut it. The people spending money on them aren't mostly dumb.
I love self-hosting stuff and even have a bias towards it, but the cost/time tradeoff is more complex than most people think.
Every company beyond a particular size surely? For many small and medium sized companies hiring an infrastructure team makes just as little sense as hiring kitchen staff to make lunch.
For medium sized companies you need "devops engineers". And in all honesty, more than you'd need sysadmins for the same deployment.
For large companies, they split up AWS responsibilities into entire departments of team (for example, all clouds have math auth so damn difficult most large companies have -not 1- but multiple departments just dealing with authorization, before you so much as start your first app)
Local reproducibility is easier, and performance is often much better
> The "cloud" reducing staff costs
Both can be true at the same time.
Also:
> Otherwise you're waking up at 3am no matter what.
Do you account for frequency and variety of wakeups here?
Yes. In my career I've dealt with way more failures due to unnecessary distributed systems (that could have been one big bare-metal box) rather than hardware failures.
You can never eliminate wake-ups, but I find bare-metal systems to have much less moving parts means you eliminate a whole bunch of failure scenarios so you're only left with actual hardware failure (and HW is pretty reliable nowadays).
There was, I have to admit, a log message that explained the problem... once I could find the specific log message and understand the 45 steps in the chain that got to that spot.
This doesn’t make sense as an argument. The reason the cloud is more complex is because that complexity is available. Under a certain size, a large number of cloud products simply can’t be managed in-house (and certainly not altogether).
Also your argument is incorrect in my experience.
At a smaller business I worked at, I was able to use these services to achieve uptime and performance that I couldn’t achieve self-hosted, because I had to spend time on the product itself. So yeah, we’d saved on infrastructure engineers.
At larger scales, what your false dichotomy suggests also doesn’t actually happen. Where I work now, our data stores are all self-managed on top of EC2/Azure, where performance and reliability are critical. But we don’t self-host everything. For example, we use SES to send our emails and we use RDS for our app DB, because their performance profiles and uptime guarantees are more than acceptable for the price we pay. That frees up our platform engineers to spend their energy on keeping our uptime on our critical services.
Just be careful not to accept more complexity just because it is available, which is what the AWS evangelists often try to sell. After all, we should always make an informed decision when adding a new dependency, whether in code or in infrastructure.
How sure are you about that one? All of my hetzner vm`s reach an uptime if 99.9% something.
I could see more then one small business stack fitting onto a single of those vm`s.
Most projects I have worked on in my career have never seen more than a hundred concurrent users. If something goes down on Saturday, I am going to fix it on Monday.
I have worked on internal tools were I just added a postgres DB to the docker setup and that was it. 5 Minute of work and no issues at all. Sure if you have something customer facing, you need to do a bit more and setup a good backup strategy but that really isn't magic.
Can we honestly say that cloud services taking a half hour to two hours a month of someone's time on average is completely unheard of?
Hiring and replacing engineers who can and want to manage database servers can be hard or scary for employers.
However a self-hosted PostgreSQL on a bare metal server with NVMe SSDs will much faster than what RDS is capable of. Especially at the same price points.
> Standard Postgres compiled with some AWS-specific monitoring hooks
… and other operational tools deployed alongside it. That’s not always true: RDS classic may be those things, but RDS Aurora/Serverless is anything but.
As to whether
> self-hosted PostgreSQL on a bare metal server with NVMe SSDs will much faster than what RDS is capable of
That’s often but not always true. Plenty of workloads will perform better on RDS (read auto scaling is huge in Serverless: you can have new read replica nodes auto-launch in response to e.g. a wave of concurrent, massive reporting queries; many queries can benefit from RDS’s additions to/modifications of the pg buffer cache system that work with the underlying storage)—and that’s even with the VM tax and the networked-storage tax! Of course, it’ll cost more in real money whether or not it performs better, further complicating the cost/benefit analysis here.
Also, pedantically, you can run RDS on bare metal with local NVMEs.
Only if you like your data to evaporate when the server stops.
I'm relatively sure that the processing power and memory you can buy on OVH / Hetzner / co. is larger and cheaper even if you take into account peaks in your usage patterns.
Later I ran a v2 of that service on k8s. The architecture also changed a lot, hosting many smaller servers sharing the same psql server(Not really microservice-related, think more "collective of smaller services ran by different people"). I have hit some issues relating to maxing out the max connections, but that's about it.
This is something I do on my free time so SLA isn't an issue, meaning I've had the ability to learn the ropes of running PSQL without many bad consequences. I'm really happy I have had this opportunity.
My conclusion is that running PSQL is totally fine if you just set up proper monitoring. If you are an engineer that works with infrastructure, even just because nobody else can/wants to, hosting PSQL is probably fine for you. Just RTFM.
Sometimes it is nice to simplify the conversation with non-tech management. Oh, you want HA / DR / etc? We click a button and you get it (multi-AZ). Clicking the button doubles your DB costs from x to y. Please choose.
Then you have one less repeating conversation and someone to blame.
Managed database services mostly automate a subset of routine operational work, things like backups, some configuration management, and software upgrades. But they don't remove the need for real database operations. You still have to validate restores, build and rehearse a disaster recovery plan, design and review schemas, review and optimize queries, tune indexes, and fine-tune configuration, among other essentials.
In one incident, AWS support couldn't determine what was wrong with an RDS cluster and advised us to "try restarting it".
Bottom line: even with managed databases, you still need people on the team who are strong in DBOps. You need standard operating procedures and automation, built by your team. Without that expertise, you're taking on serious risk, including potentially catastrophic failure modes.
I would've very much preferred being able to SSH in myself and fix it on the spot. Ironically the only reason it ran out of space in the first place is that the AWS markup on that is so huge we were operating with little margin for error; none of that would happen with a bare-metal host where I can rent 1TB of NVME for a mere 20 bucks a month.
As far as I know we never got any kind of compensation for this, so RDS ended up being a net negative for this company, tens of thousands spent over a few years for laptop-grade performance and it couldn't even do its promised job the only time it was needed.
And I really recommend starting with *default* k3s, do not look at any alternatives to cni, csi, networked storage - treat your first cluster as something that can spontaniously fail and don't bother keeping it clean learn as much as you can.
Once you have that, you can use great open-source k8s native controllers which take care of vast majority of requirements when it comes to self-hosting and save more time in the long run than it took to set up and learn these things.
Honerable mentions: k9s, lens(I do not suggest using it in the long-term, but UI is really good as a starting point), rancher webui.
PostgreSQL specifically: https://github.com/cloudnative-pg/cloudnative-pg If you really want networked storage: https://github.com/longhorn/longhorn
I do not recommend ceph unless you are okay with not using shared filesystems as they have a bunch of gotchas or if you want S3 without having to install a dedicated deployment for it.
As someone who has operated Postgres clusters for over a decade before k8s was even a thing, I fully recommend just using a Postgres operator like this one and moving on. The out of box config is sane, it’s easy to override things, and failover/etc has been working flawlessly for years. It’s just the right line between total DIY and the simplicity of having a hosted solution. Postgres is solved, next problem.
Overall, a good experience. Very stable service and when performance issues did periodically arise, I like that we had full access to all details to understand the root cause and tune details.
Nobody was employeed as a full-time DBA. We had plenty of other things going on in addition to running PostgreSQL.
A blog post that went into the details would be awesome. I know Postgres has some docs for this (https://www.postgresql.org/docs/current/backup.html), but it's too theoretical. I want to see a one-stop-shop with everything you'd reasonably need to know to self host: like monitoring uptime, backups, stuff like that.
I did this for just under two years, and I've lost count of how many times one or more of the nodes went down and I had to manually deregister it from the cluster with repmgr, clone a new vm and promote a healthy node to primary. I ended up writing an internal wiki page with the steps. I never got it: if one of the purposes of clusters is having higher availability, why did repmgr not handle zombie primaries?
Again, I'm probably just an idiot out of my depth with this. And I probably didn't need a cluster anyway, although with the nodes failing like they did, I didn't feel comfortable moving to a single node setup as well.
I eventually switched to managed postgres, and it's amazing being able to file a sev1 for someone else to handle when things go down, instead of the responsibility being on me.
For client projects, however, I always try and sell them on paying the AWS fees, simply because it shifts the responsibility of the hardware being "up" to someone else. It does not inherently solve the downtime problem, but it allows me to say, "we'll have to wait until they've sorted this out, Ikea and Disney are down, too."
Doesn't always work like that and isn't always a tried-and-true excuse, but generally lets me sleep much better at night.
With limited budgets, however, it's hard to accept the cost of RDS (and we're talking with at least one staging environment) when comparing it to a very tight 3-node Galera cluster running on Hetzner at barely a couple of bucks a month.
Or Cloudflare, titan at the front, being down again today and the past two days (intermittently) after also being down a few weeks ago and earlier this year as well. Also had SQS queues time out several times this week, they picked up again shortly, but it's not like those things ...never happen on managed environments. They happen quite a bit.
It is. You need to answer the question: what are the changes of your service being down for lets say 4 hours or some security patch isn't properly applied or you have not followed the best practices in terms of security? Many people are technically unable, lack the time or the resources to be able to confidently address that question, hence paying for someone else to do it.
Your time is money though. You are saving money but giving up time.
Like everything, it is always cheaper to do it (it being cooking at home, cleaning your home, fixing your own car, etc) yourself (if you don't include the cost of your own time doing the service you normally pay someone else for).
You see the issue?
Like, I’m all for not procuring things that it makes more sense to own/build (and I know most businesses have piss-poor instincts on which is which—hell, I work for the government! I can see firsthand the consequences of outsourcing decision making to contractors, rather than just outsourcing implementation).
But it’s very case-by-case. There’s no general rule like “always prefer self hosting” or “always rent real estate, never buy” that applies broadly enough to be useful.
I have never, ever, ever had a SQL box go down. I've had a web server go down once. I had someone who probably shouldn't have had access to a server accidentally turn one off once.
The only major outage I've had (2/3 hours) was when the box was also self-hosting an email server and I accidentally caused it to flood itself with failed delivery notices with a deploy.
I may have cried a little in frustration and panic but it got fixed in the end.
I actually find using cloud hosted SQL in some ways harder and more complicated because it's such a confusing mess of cost and what you're actually getting. The only big complication is setting up backups, and that's a one-off task.
Obviously it depends on the operational overhead of specific technology.
Standard Postgres compiled with some AWS-specific monitoring hooks
A custom backup system using EBS snapshots
Automated configuration management via Chef/Puppet/Ansible
Load balancers and connection pooling (PgBouncer)
Monitoring integration with CloudWatch
Automated failover scripting
The problem i find with RDS (and most other managed Postgres) is that they limit your options for how you want to design your database architecture. For instance, if write consistency is important to you want to support synchronous replication, there is no way to do this in RDS without either Aurora or having the readers in another AZ. The other issue is that you only have access to logical replication, because you don't have access to your WAL archive, so it makes moving off RDS much more difficult.
Is this actually the "common" view (in this context)?
I've got decades with databases so I cannot even begin to fathom where such an attitude would develop, but, is it?
Boggling.
This is when you use SQLite, not Postgres. Easy enough to turn into Postgres later, nothing to set up. It already works. And backups are literally just "it's a file, incremental backup by your daily backups already covers this".
arichard123•2h ago
newsoftheday•1h ago
moxplod•49m ago