I started using it in late 2023, and never paid anything for it, so I definitely should not complain. Two and half years of mostly loyal service, but sometimes less is more, and I am moving on.
Background
Oracle Cloud offers very generous free tier. You could basically get (if you jump through some hoops to fight for sparse capacity) 8 VCPU/24GB RAM/200GB ARM machine, or consistently 2 VCPUs of X86 with also 200GB storage.
The ARM experiment (or lack of it) in late 2023
Initially, after retrying tofu apply many, many times, I got an ARM machine. Then I did not use its cores or memory enough, and it got deprovisioned perhaps month or so later. After that, I could not get easily new node, due to capacity issues, and apparently it has been like that ever since (supposedly if you switch to PAYG customer, you get preference even over free resource allocations, but I never tried that). Anyway, as I do not really lack compute capacity, that was not a priority to address.
X86 nodes
The two x86 single VCPU nodes that I used ever since were very, very slow CPU wise (fraction of of real core each), but relatively reliable. Initially I started with Ubuntu 22.04 and later on migrated them to Ubuntu 24.04, and they served faithfully(?) as
oraakkeli.fingon.iki.fi- my VPN exit node (initially IPsec, later on Tailscale)www.fingon.iki.fi- the fancy website you are browsing
I used them to experiment with (not full list)
- different VPN technologies
- different provisioning tools (pyinfra, terraform, opentofu, pulumi)
Of course, while the constant allocation was 2 nodes, the actual nodes were recreated many, many times as I played with different provisioning solutions for them.
What broke the camels back yesterday?
I noticed that my Tailscale exit node was not available last week when I was travelling, so I decided to investigate what was up today. As it turns out, both of the nodes disappeared on 13th of May, 2026, without trace, around 18:40 (metrics) and 18:00 (logs). I tried to log in to the cloud console, and got this:
I did reset password flow (I am not sure what year I last logged in there - probably some time in 2025), and the outcome was same. As reset password flow worked fine, I assume my account got just flipped to some silent suspend mode which is not mentioned in the UI (there are some Internet references in which people get message about suspension - this does not seem to happen here).
To test the silent treatment hypothesis, I tested different browsers (same result), as well as my dead fingon Oracle cloud tenant (I managed to lose FIDO token to it at some point, and I have no other way of recovering the account without talking to support, which I would rather not do). The reset password flow worked fine on the previous account, although of course I could not supply the 2FA so login stalled there.
So I assume that tenant is simply gone for me.
Moving on
This is not much of a loss - I have been mostly self hosting years, and while this website has been occasionally hosted at Kapsi (a backup of it still is, and will be, at https://fingon.kapsi.fi), but as the reliability of the web server there was bit bad compared to Oracle, it had been since October 2024 at Oracle.
Here are the monitoring results for completeness - kapsi version of homepage (failed probes from frankenrouter):
Oracle page (I think 2026 monitoring is bugged, as I did not get alert when Oracle killed the nodes):
As my ISP is not flawless, and my home infra sometimes is rocky, the source of probes is not 100% reliable either. But the Oracle Cloud nodes did serve better over time, and I am glad for their service. But now it is time for git rm of some pyinfra code, as I do not leave old stuff hanging.
[main c7341ce] oraakkeli,www: He's dead, Jim.
33 files changed, 17 insertions(+), 1336 deletions(-)
I’ll perhaps need to set up a new Tailscale exit node somewhere, at some point, but otherwise they won’t be missed.