TierHive General thread | Discuss, Updates, questions | LATEST: 8GB Ram available in some locations

flips

Thanks @AnthonySmith … And I thought Gentoo was lightweight …

AnthonySmith

@flips said:
Thanks @AnthonySmith … And I thought Gentoo was lightweight …

Alpine is the option now for minimal

Not_Oles

@AnthonySmith said:
I will try and sort a Gentoo template next week.

On the beach near here sometimes there are a lot of flies.

Two of the flies just buzzed past my Chromebook!

One fly said, "Gentoo next week!"

The second fly replied, "NetBSD too!"

Bzzzt! Right in your ear! . . . Okay, I will shut up now.

flips

... And then GNU/Hurd the week after?

Not_Oles

@flips It doesn't matter, since from me, but . . . you have been hurd! /s

AnthonySmith

@Not_Oles said:

@AnthonySmith said:
I will try and sort a Gentoo template next week.

On the beach near here sometimes there are a lot of flies.

Two of the flies just buzzed past my Chromebook!

One fly said, "Gentoo next week!"

The second fly replied, "NetBSD too!"

Bzzzt! Right in your ear! . . . Okay, I will shut up now.

@WSS basically threatened my life if I spend any time on NetBSD

Gentoo already maintain a cloud init compatible template so I am hoping I only need an hour or so to sort it out.

nameserver

I have been spending the last two days playing with the system.

Trying to do HAProxy with geo dns, but couldn't manage to get it to work.

IPv6 has been working, but I couldn't manage to get ports other than 22 to open.

Came across two times of the same issue that when booting up the VPS, it failed to start as it seems to not be able to get an ip address. Deleting and recreating a VPS works, reinstalling doesn't.

I noticed with the same config, in different location, the CPU performance and network differs, I assume it's the different performance of the server?

nameserver

@nameserver said: IPv6 has been working, but I couldn't manage to get ports other than 22 to open.

Edit: all ports except 53 is working. Somehow port 53 is blocked.

Not_Oles

More use cases for an existing feature. . . .

I like the reinstall page because it shows which OSes are available and which OSes are not available given the existing RAM and disk configuration.

On the Create New VPS page, there is only the --Select OS-- drop down dialog, which shows only a list of OSes and doesn't show which OSes are available for the currently configured RAM and disk configuration.

On my Chromebook, the --Seect OS-- listing shows as almost invisibly light gray on bright white background. Additionally, my Chromebook refuses to take a screenshot showing the listing. If I take the screenshot with the listing showing, the listing does not appear in the screenshot, which somehow shows the page as it was before the drop down was activated:

Maybe it might be nice if the informative Reinstall page could be integrated into the Create New VPS process, replacing the drop down? Thanks for considering!

---

So a Chromebook screenshot somehow isn't really a screenshot? I am tempted to take a picture of my Chromebook screen with my phone. Does anyone know why my Chromebook won't do a screenshot which shows the open drop-down dialog? Thanks!

david3

It's similar taking screenshots with wayland. Any drop-downs disappear (it's not just tierhive). I think it has something to do with the way the display system works.

AnthonySmith

@Not_Oles said: Maybe it might be nice if the informative Reinstall page could be integrated into the Create New VPS process, replacing the drop down? Thanks for considering!

yeah I was thinking about that, I have some notes, currently when you select an OS it auto sets you to the minimum resources for that OS, what I really need to do though, well not me but TierHive needs to do, is provide some sort of live update on which location can create based on what you select rather than just stopping you after, because thats annoying.

During testing, obviously, we did not have high contention on resources, so that issue never came up.

AnthonySmith

First unplanned significant outage today took a UK node but impacted 2 and IPv6 and control panel functions on the entire platform.

Fun way to start a Sunday.

This is why we are in Alpha , and it highlighted some points of fragility with the system, as always, due to scale not matching testing.

The good news is that after the impacted node was power cycled, the mesh rebuilt itself, the firewall rebuilt itself, the TierHive specific services rebuilt themselves, and the hypervisor automatically put itself into 0 priority mode, which for us means you can't create anything on it while it's fixing its own issues.

What failed was that the NAT rebuild hit a race condition due to scale, and so although virtual servers were up they were not reachable, and this also stopped the IPv6 routers from self-configuring fully, which in turn took out all of UK IPv6 on the mesh/tunnels instead of routing through France as intended for the UK if that happens.

So we need to add some timers and lock files to force things to happen in the right order. I am fairly confident that had that been the case, we would have had a full self-repair.

But the worst issue is that we have a bug in the workers' system (We don't use Laravel because urgh.. Laravel, but borrowed some principles) if a hypervisor is down for long enough, it can cause a block abd cascade failure so all workers fully time out and even our heartbeat cron cant bring them back to life, the net result is that every function just about in the whole panel does nothing when you click it as an end user, luckily this was only the case for about 30 minutes we think, but thats todays priority instead of sunday things.

But, I am afraid if you clicked reboot 200 times... your VPS is now going to reboot 200 times an hour later.

Normal service will be resumed asap.

Remember... ALPHA

nameserver

@nameserver said:

@nameserver said: IPv6 has been working, but I couldn't manage to get ports other than 22 to open.

Edit: all ports except 53 is working. Somehow port 53 is blocked.

Update, there is some serious issue with networking. It wasn't port being blocked, it was the VPS unable to reach the ipv6 gateway. Then I tried reinstalling all my VPS, but all of them failed to get an ip, even the internal ipv4 it seems, that I cannot connect to them. The issue is same in all 8 locations, no ipv6 issue in AU since it doesn't have ipv6. Tried opening ticket but support says they cannot reproduce the issue.

AnthonySmith

@nameserver said:

@nameserver said:

@nameserver said: IPv6 has been working, but I couldn't manage to get ports other than 22 to open.

Edit: all ports except 53 is working. Somehow port 53 is blocked.

Update, there is some serious issue with networking. It wasn't port being blocked, it was the VPS unable to reach the ipv6 gateway. Then I tried reinstalling all my VPS, but all of them failed to get an ip, even the internal ipv4 it seems, that I cannot connect to them. The issue is same in all 8 locations, no ipv6 issue in AU since it doesn't have ipv6. Tried opening ticket but support says they cannot reproduce the issue.

In the ticket you opened you showed a screenshot showing your VPS had a 169.254 address so whatever you are doing in your config is what is breaking the network stack. as per my answer in the ticket.

Sorry if you have updated it since there is a backlog with todays outage, there have been no other reports of similar issues and looking at my colleguues admin notes when he first replied, he setup a test VPS on your subnet, checked IPv4, checking IPv6, concluded it was a local user config issue and rmoved it and give you that reply.

I am up to my neck right now but as soon as I am finished I will do the same to validate.

AnthonySmith

Last update as it is believed everything is restored 100% now.

The issue that was causing VPS creation to fail or fail to issue an IP correctly was on us. As a result of the outage, we found the reason for the cascade failure and fixed it.

IPv6 worker was impacted in this, so IPs were not being whitelisted, you will need to remove and re-add them from the panel if this is still impacting you.

The UK VPS on the impacted nodes last state was unknown after the forced reboot of the UK node so VPS are all just left in a stopped, you can start them whenever you want.

back to business as usual.

Any more problems please open a ticket.

nameserver

@AnthonySmith said: whatever you are doing in your config

I was trying to reinstall using the control panel. Nothing else. Before that I was trying to config IPv6 using the guide provided. Only external software I installed is iptables, so would this be the case?

I'll update more details in the ticket.

Thanks for the reply.

AnthonySmith

@nameserver said:

@AnthonySmith said: whatever you are doing in your config

I was trying to reinstall using the control panel. Nothing else. Before that I was trying to config IPv6 using the guide provided. Only external software I installed is iptables, so would this be the case?

I'll update more details in the ticket.

Thanks for the reply.

probably worth reading my last update here first.

nameserver

@AnthonySmith said:

@nameserver said:

@AnthonySmith said: whatever you are doing in your config

I was trying to reinstall using the control panel. Nothing else. Before that I was trying to config IPv6 using the guide provided. Only external software I installed is iptables, so would this be the case?

I'll update more details in the ticket.

Thanks for the reply.

probably worth reading my last update here first.

Sorry, didn't see the message as I was sending my reply. Thanks for the help.

WSS

My IPv6 isn't working in Vint Hill.

$ ping 2a11:6c7:2300:9508::1
PING 2a11:6c7:2300:9508::1 (2a11:6c7:2300:9508::1) 56 data bytes
--- 2a11:6c7:2300:9508::1 ping statistics ---
68 packets transmitted, 0 received, 100% packet loss, time 68615ms

---

For whatever reason I had to give it a hard reboot. It just would not talk to the gateway anymore.

Not_Oles

My Vint Hill IPv6 also was out. It came back after removing and re-adding the instance's IPv6 address in the Control Panel as advised in the notice seen in the Control Panel[1] for UK instances.

friendly:~# date
Sun May 17 18:03:55 UTC 2026
friendly:~# ping6 -c 2 ipv6.google.com
PING ipv6.google.com (2607:f8b0:4023:1006::66): 56 data bytes
64 bytes from 2607:f8b0:4023:1006::66: seq=0 ttl=103 time=546.922 ms
64 bytes from 2607:f8b0:4023:1006::66: seq=1 ttl=103 time=33.393 ms

--- ipv6.google.com ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss
round-trip min/avg/max = 33.393/290.157/546.922 ms
friendly:~# 

[1] UK Partial Outage
We are aware of the partial outage in the UK. We hope to have this resolved today. Sorry for the inconvenience. - - Update: All issues should not [now?] be 100% resolved, if you created a VPS that is acting strange in the last 6 hours you may need to recreate it, if IPv6 is not working, please remove and re-add the individual address in the panel to re-whitelist them.

Not_Oles

Wow! I got a Tierhive instance in KC! Thanks @AnthonySmith!

kc:~# date
Sun May 17 18:44:57 UTC 2026
kc:~# uptime
 18:45:00 up 2 min,  0 users,  load average: 2.54, 2.00, 0.83
kc:~# uname -a
Linux kc.tommiller.net 6.18.31-0-virt #1-Alpine SMP PREEMPT_DYNAMIC 2026-05-16 07:00:20 x86_64 Linux
kc:~# 

Both IPv4 and IPv6 seem to work. While connected via ssh over IPv4:

kc:~# ping6 -c 2 ipv6.google.com
PING ipv6.google.com (2607:f8b0:4009:80c::200e): 56 data bytes
64 bytes from 2607:f8b0:4009:80c::200e: seq=0 ttl=116 time=114.246 ms
64 bytes from 2607:f8b0:4009:80c::200e: seq=1 ttl=116 time=25.106 ms

--- ipv6.google.com ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss
round-trip min/avg/max = 25.106/69.676/114.246 ms
kc:~# 

AnthonySmith

We built the ipv6 mesh to be a bit to interdependent so the framework for elastic IPs would be there.

However it does depend on some traffic being active not to make bad decisions with all the noise Ipv6 generates, it's actually now down to 1.3% use and less than 1% of all traffic.

It's at the point I am artificially generating traffic between ipv6 routers to keep the automation from making bad choices.

Starting to feel like I wasted a month of 2 people's time, it's the NetBSD of internet protocols.

WSS

@AnthonySmith said: it's the NetBSD of internet protocols.

Bruh.

WSS

BRB. Gonna pingflood the gateway to keep it open.

nameserver

I have been struggling for the past few hours trying to get iptables with ipv6 to work. Setup the same rule to forward ipv6 traffic for port 80, 443 TCP, 53 TCP and UDP, but port 53 never works. I change the port number to something like 5353, works, same setup.

Is port 53 blocked on the network by any chance?

Not_Oles

Is there a good benchmark script which might run on TierHive's Alpine 3.23 256 MB RAM, 1 GB disk VPS?

Perhaps nench ("new bench")?

Below are Kansas City TierHive node nench results from running a local copy of nench.sh after apk update, upgrade to Alpine 3.23.4, and with locally installed bash, coreutils, ioping, openssl, and util-linux. The CPU Priority, IOPS Tier, and Network Tier all were set to Medium. A 64 MB swap file also was added.

-------------------------------------------------
 nench.sh v2019.07.20 -- https://git.io/nench.sh
 benchmark timestamp:    2026-05-18 04:28:36 UTC
-------------------------------------------------

Processor:    AMD EPYC 7B13 64-Core Processor
CPU cores:    1
Frequency:    2250.000 MHz
RAM:          217.0M
Swap:         64.0M
Kernel:       Linux 6.18.31-0-virt x86_64

Disks:
vda      1G  HDD

CPU: SHA256-hashing 500 MB
    3.534 seconds
CPU: bzip2-compressing 500 MB
    59.240 seconds
CPU: AES-encrypting 500 MB
    7.111 seconds

ioping: seek rate
    min/avg/max/mdev = 469.6 us / 6.82 ms / 99.4 ms / 8.49 ms
ioping: sequential read speed
    generated 263 requests in 5.00 s, 65.8 MiB, 52 iops, 13.1 MiB/s

dd: sequential write speed
    1st run:    14.88 MiB/s
    2nd run:    27.75 MiB/s
    3rd run:    57.32 MiB/s
    average:    33.32 MiB/s

IPv4 speedtests
    your IPv4:    208.84.xxx.xxx

    Cachefly CDN:         7.28 MiB/s
    Leaseweb (NL):        0.02 MiB/s
    Softlayer DAL (US):   0.00 MiB/s
    Online.net (FR):      2.64 MiB/s
    OVH BHS (CA):         4.97 MiB/s

IPv6 speedtests
    your IPv6:    2a11:6c7:xxxx:xxxx

    Leaseweb (NL):        0.03 MiB/s
    Softlayer DAL (US):   0.00 MiB/s
    Online.net (FR):      0.93 MiB/s
    OVH BHS (CA):         1.89 MiB/s
-------------------------------------------------

AnthonySmith

@Not_Oles said: Is there a good benchmark script which might run on TierHive's Alpine 3.23 256 MB RAM, 1 GB disk VPS?

I would recommend low expectations and human intuition; it takes up no disk space.

@nameserver said: Is port 53 blocked on the network by any chance?

Yes, sorry, I just dont want to deal with amplification attacks, you might know what you are doing, the other 100 people that copy and paste something do not.

AnthonySmith

Adding extra capacity to Vint Hill, I will get in touch with @Advin this week to increase capacity in Kansas, as it's running like shit now; he replied to my last request, and I missed his reply.

Vint Hill is also getting HDD capacity so cheap HDD will be available in the USA as well as a backup target becoming available outside of london, I plan to do the same in Singapore.

No one uses Australia... someone in Australia please tell me why

AnthonySmith

IPv6 bug found, it should be more stable (and self-healing) within the next day or so, testing is ongoing in the sandpit now

nameserver

@AnthonySmith said:

@nameserver said: Is port 53 blocked on the network by any chance?

Yes, sorry, I just dont want to deal with amplification attacks, you might know what you are doing, the other 100 people that copy and paste something do not.

Ahh. Spent the whole weekend trying to figure out why my port forwarding was not working. Also just to let you know that in FR port 53 is not blocked.

Not_Oles

@AnthonySmith said: @Not_Oles said: Is there a good benchmark script which might run on TierHive's Alpine 3.23 256 MB RAM, 1 GB disk VPS?

I would recommend low expectations and human intuition; it takes up no disk space.

Thanks @AnthonySmith! I replaced[1] nench.sh's no-longer available Softlayer 4 and 6 speed tests with Hetzner in Ashburn.

Here is another run:

-------------------------------------------------
 nench.sh v2019.07.20 -- https://git.io/nench.sh
 benchmark timestamp:    2026-05-18 18:44:52 UTC
-------------------------------------------------

Processor:    AMD EPYC 7B13 64-Core Processor
CPU cores:    1
Frequency:    2250.000 MHz
RAM:          217.0M
Swap:         64.0M
Kernel:       Linux 6.18.31-0-virt x86_64

Disks:
vda      1G  HDD

CPU: SHA256-hashing 500 MB
    3.175 seconds
CPU: bzip2-compressing 500 MB
    58.308 seconds
CPU: AES-encrypting 500 MB
    10.255 seconds

ioping: seek rate
    min/avg/max/mdev = 907.0 us / 11.9 ms / 83.9 ms / 10.5 ms
ioping: sequential read speed
    generated 440 requests in 5.00 s, 110 MiB, 87 iops, 22.0 MiB/s

dd: sequential write speed
    1st run:    63.42 MiB/s
    2nd run:    50.64 MiB/s
    3rd run:    77.72 MiB/s
    average:    63.93 MiB/s

IPv4 speedtests
    your IPv4:    208.84.xxx.xxx

    Cachefly CDN:         5.87 MiB/s
    Hetzner ASH (US):     2.81 MiB/s
    Leaseweb (NL):        0.03 MiB/s
    Online.net (FR):      4.02 MiB/s
    OVH BHS (CA):         6.20 MiB/s

IPv6 speedtests
    your IPv6:    2a11:6c7:xxxx:xxxx

    Hetzner ASH (US):     2.42 MiB/s
    Leaseweb (NL):        0.03 MiB/s
    Online.net (FR):      3.46 MiB/s
    OVH BHS (CA):         1.38 MiB/s
-------------------------------------------------

kc:~$ 

[1]

kc:~$ diff nench.sh~ nench.sh
--- nench.sh~
+++ nench.sh
@@ -259,13 +259,17 @@
     download_benchmark -4 http://cachefly.cachefly.net/100mb.test | \
         Bps_to_MiBps
 
+    printf '    Hetzner ASH (US):     '
+    download_benchmark -4 https://ash-speed.hetzner.com/100MB.bin | \
+        Bps_to_MiBps
+
     printf '    Leaseweb (NL):        '
     download_benchmark -4 http://mirror.nl.leaseweb.net/speedtest/100mb.bin | \
         Bps_to_MiBps
 
-    printf '    Softlayer DAL (US):   '
-    download_benchmark -4 http://speedtest.dal06.softlayer.com/downloads/test100.zip | \
-        Bps_to_MiBps
+#   printf '    Softlayer DAL (US):   '
+#   download_benchmark -4 http://speedtest.dal06.softlayer.com/downloads/test100.zip | \
+#       Bps_to_MiBps
 
     printf '    Online.net (FR):      '
     download_benchmark -4 http://ping.online.net/100Mo.dat | \
@@ -287,14 +291,18 @@
     printf 'IPv6 speedtests\n'
     printf '    your IPv6:    %s\n' "$(redact_ip "$ipv6")"
     printf '\n'
+  
+    printf '    Hetzner ASH (US):     '
+    download_benchmark -6 https://ash-speed.hetzner.com/100MB.bin | \
+        Bps_to_MiBps
 
     printf '    Leaseweb (NL):        '
     download_benchmark -6 http://mirror.nl.leaseweb.net/speedtest/100mb.bin | \
         Bps_to_MiBps
 
-    printf '    Softlayer DAL (US):   '
-    download_benchmark -6 http://speedtest.dal06.softlayer.com/downloads/test100.zip | \
-        Bps_to_MiBps
+#    printf '    Softlayer DAL (US):   '
+#    download_benchmark -6 http://speedtest.dal06.softlayer.com/downloads/test100.zip | \
+#       Bps_to_MiBps
 
     printf '    Online.net (FR):      '
     download_benchmark -6 http://ping6.online.net/100Mo.dat | \
kc:~$ 

Not_Oles

@AnthonySmith said:

with all the noise Ipv6 generates, it's actually now down to 1.3% use and less than 1% of all traffic.

It's at the point I am artificially generating traffic between ipv6 routers to keep the automation from making bad choices.

Starting to feel like I wasted a month of 2 people's time, it's the NetBSD of internet protocols.

Thanks for mentioning NetBSD! I'm still confused about how best to install NetBSD on TierHive. Meanwhile, IPv6 is working on tiny TierHive Alpine VPSes, and, maybe, their small use of IPv6 will help TierHive's IPv6 automation keep out of trouble.

Always best wishes! And Thanks!

cmeerw

@Not_Oles said:

@AnthonySmith said:

with all the noise Ipv6 generates, it's actually now down to 1.3% use and less than 1% of all traffic.

It's at the point I am artificially generating traffic between ipv6 routers to keep the automation from making bad choices.

Starting to feel like I wasted a month of 2 people's time, it's the NetBSD of internet protocols.

I'm still confused about how best to install NetBSD on TierHive.

Basically

• boot into an Alpine image via netboot (or something similar)
• write the NetBSD image to the disk and reboot
• in the NetBSD boot loader, interrupt it, set the vesa mode to something like 0x101 (vesa 0x101) and boot into NetBSD (boot netbsd)
• if you are using one of my images, it will reboot again after resizing the root partition, so you have to do the vesa mode setting again.
• login as root (via the VNC console), set users/passwords and networking, you'll also want to add the vesa setting to NetBSD's boot config

No idea why that vesa mode setting is needed on TierHive (never seen this issue on other VPSes)

AnthonySmith

IPv6 should be a lot more stable now. We also added a consistency cron for ipv6 just in case, its been running a few hours now since we put the fixes in place, and no more issues have been detected or fixes required.

Thanks for the patience and reports.

luis

I've just found this out, really awesome project! Definitely giving it a try.
Am I being thick or is there no way to add slow (HDD) storage when creating a VPS?

AnthonySmith

@luis said:
I've just found this out, really awesome project! Definitely giving it a try.
Am I being thick or is there no way to add slow (HDD) storage when creating a VPS?

No sorry it was disabled, try again, pick London (only location with HDD capacity right now)

Should be adding USA storage later today or tomorrow.

luis

@AnthonySmith said:

@luis said:
I've just found this out, really awesome project! Definitely giving it a try.
Am I being thick or is there no way to add slow (HDD) storage when creating a VPS?

No sorry it was disabled, try again, pick London (only location with HDD capacity right now)

Should be adding USA storage later today or tomorrow.

I'm just checking out prices at the moment, are you planning to bump these in the future?
2.05$ a month for 2GB RAM / 10GB SSD / 200GB HDD seems quite attractive!

AnthonySmith

@luis said:

@AnthonySmith said:

@luis said:
I've just found this out, really awesome project! Definitely giving it a try.
Am I being thick or is there no way to add slow (HDD) storage when creating a VPS?

No sorry it was disabled, try again, pick London (only location with HDD capacity right now)

Should be adding USA storage later today or tomorrow.

I'm just checking out prices at the moment, are you planning to bump these in the future?
2.05$ a month for 2GB RAM / 10GB SSD / 200GB HDD seems quite attractive!

not planning on it no, but you know, the world is a crazy place right now, if our costs increased significantly, we would not have any choice, but the goal is to actually reduce pricing if anything.

The only thing I would add to that is that we are in Alpha, we pretty much guessed at pricing on some things not being able to predict use, so if we have to change prices, we will, if we can reduce them, we will, in the sandpit version, we already charge 50% less for powered-off virtual servers. That will make it into live fairly soon and the only changes we have made so far have been price reductions.

Also some regions are going to be expensive, we may have to add a multiplier for specific locations.

Not_Oles

Friendly greetings!

Here's the netboot.xyz rescue system loaded at TierHIve:

Here is what is shown after clicking BSD installs . . . some BSDs might seem to be missing? Where is NetBSD? Thanks!

Not_Oles

@Not_Oles said: some BSDs might seem to be missing? Where is NetBSD?

https://github.com/netbootxyz/netboot.xyz/issues/1437

WSS

Jesus Christ Tom, get a hobby.

AnthonySmith

@Not_Oles said:

@Not_Oles said: some BSDs might seem to be missing? Where is NetBSD?

https://github.com/netbootxyz/netboot.xyz/issues/1437

Not really a TierHive issue but you know netbook.xyz supports chain loading of an iso ?

Not_Oles

@AnthonySmith Yes, thanks!

I am just having fun on TierHive trying various, different, new-to-me things in various, different, new-to-me ways.

Mostly, right now, I am trying to fit what I am doing g into 1 vCPU, 256 GB RAM, and 1 GB disk -- in other words, the requirements on Tierhive for Alpine 3.23.2.

Yesterday, I had a fun afternoon trying to install OpenBSD from Tierhive's netboot.xyz BSD menu (screenshot above). What happened was that the test VPS appeared to run out of disk space during the KASLR relinking. I doubled the disk space to 2 GB and tried again with what appeared to be the same result:

Today's plan is possibly to try installing NetBSD again with these iPXE commands:

set net0/ip <YOUR_STATIC_IP>
set net0/netmask <YOUR_SUBNET_MASK>
set net0/gateway <YOUR_GATEWAY_IP>
set dns 74.82.42.42
kernel https://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/amd64/binary/kernel/netbsd-INSTALL.gz
boot

Soon, I expect to try @cmeerw's suggestion:

boot into an Alpine image via netboot (or something similar)
write the NetBSD image to the disk and reboot
in the NetBSD boot loader, interrupt it, set the vesa mode to something like 0x101 (vesa 0x101) and boot into NetBSD (boot netbsd)
if you are using one of my images, it will reboot again after resizing the root partition, so you have to do the vesa mode setting again.
login as root (via the VNC console), set users/passwords and networking, you'll also want to add the vesa setting to NetBSD's boot config

And, previously, from @cmeerw:

IIRC, I used netboot xyz (manually configuring the network interface as I didn't have DHCP enabled) to boot into a live Alpine Linux system which I only used to write the NetBSD image to the disk. Then boot from the disk (but I think you have to fiddle with the VESA mode), and then you can manually configure networking.

I also want to try other very small OSes, maybe something like Oasis, SliTaz, and more.

If anybody knows why the OpenBSD install didn't seem to work, even at 2 GB disk, I would really appreciate a hint.

Thanks again, Ant! TierHive's smallest and lightest VPSes are tons of fun!

flips

TinyCore Linux, maybe?

Not_Oles

Today's results from nench, slightly modified to substitute Hetzner ASH for the no-longer available SoftLayer. These tiny Alpine VPSes need a speed test that can run on 256 MB RAM and 1 GB disk.

VPS named "friendly" in TierHive's Vint Hill, VA USA location:

-------------------------------------------------
 nench.sh v2019.07.20 -- https://git.io/nench.sh
 benchmark timestamp:    2026-05-20 21:16:34 UTC
-------------------------------------------------

Processor:    Intel(R) Xeon(R) E-2236 CPU @ 3.40GHz
CPU cores:    1
Frequency:    3408.008 MHz
RAM:          217.0M
Swap:         -
Kernel:       Linux 6.18.32-1-virt x86_64

Disks:
vda      1G  HDD

CPU: SHA256-hashing 500 MB
    1.864 seconds
CPU: bzip2-compressing 500 MB
    8.664 seconds
CPU: AES-encrypting 500 MB
    1.287 seconds

ioping: seek rate
    min/avg/max/mdev = 192.4 us / 3.99 ms / 19.8 ms / 1.59 ms
ioping: sequential read speed
    generated 1.26 k requests in 5.00 s, 315 MiB, 251 iops, 63.0 MiB/s

dd: sequential write speed
    1st run:    414.85 MiB/s
    2nd run:    284.19 MiB/s
    3rd run:    170.71 MiB/s
    average:    289.92 MiB/s

IPv4 speedtests
    your IPv4:    147.135.xxx.xxx

    Cachefly CDN:         17.38 MiB/s
    Hetzner ASH (US):     31.29 MiB/s
    Leaseweb (NL):        0.05 MiB/s
    Online.net (FR):      10.08 MiB/s
    OVH BHS (CA):         25.37 MiB/s

IPv6 speedtests
    your IPv6:    2a11:6c7:xxxx::xxxx

    Hetzner ASH (US):     12.42 MiB/s
    Leaseweb (NL):        0.01 MiB/s
    Online.net (FR):      9.42 MiB/s
    OVH BHS (CA):         9.72 MiB/s
-------------------------------------------------

friendly:~$ 

Kansas City, MO USA VPS named "kc":

-------------------------------------------------
 nench.sh v2019.07.20 -- https://git.io/nench.sh
 benchmark timestamp:    2026-05-20 21:28:57 UTC
-------------------------------------------------

Processor:    AMD EPYC 7B13 64-Core Processor
CPU cores:    1
Frequency:    2250.000 MHz
RAM:          217.0M
Swap:         64.0M
Kernel:       Linux 6.18.31-0-virt x86_64

Disks:
vda      1G  HDD

CPU: SHA256-hashing 500 MB
    5.637 seconds
CPU: bzip2-compressing 500 MB
    46.762 seconds
CPU: AES-encrypting 500 MB
    6.655 seconds

ioping: seek rate
    min/avg/max/mdev = 652.3 us / 10.1 ms / 105.5 ms / 9.73 ms
ioping: sequential read speed
    generated 540 requests in 5.00 s, 135 MiB, 107 iops, 27.0 MiB/s

dd: sequential write speed
    1st run:    11.83 MiB/s
    2nd run:    25.94 MiB/s
    3rd run:    33.09 MiB/s
    average:    23.62 MiB/s

IPv4 speedtests
    your IPv4:    208.84.xxx.xxx

    Cachefly CDN:         3.41 MiB/s
    Hetzner ASH (US):     3.52 MiB/s
    Leaseweb (NL):        0.02 MiB/s
    Online.net (FR):      4.97 MiB/s
    OVH BHS (CA):         8.30 MiB/s

IPv6 speedtests
    your IPv6:    2a11:6c7:xxxx::xxxx

    Hetzner ASH (US):     1.59 MiB/s
    Leaseweb (NL):        0.02 MiB/s
    Online.net (FR):      2.74 MiB/s
    OVH BHS (CA):         1.65 MiB/s
-------------------------------------------------

kc:~$ 

TemporiousOne

@Not_Oles said:
Today's results from nench, slightly modified to substitute Hetzner ASH for the no-longer available SoftLayer. These tiny Alpine VPSes need a speed test that can run on 256 MB RAM and 1 GB disk.

VPS named "friendly" in TierHive's Vint Hill, VA USA location:

~~~

nench.sh v2019.07.20 -- https://git.io/nench.sh

benchmark timestamp: 2026-05-20 21:16:34 UTC

Processor: Intel(R) Xeon(R) E-2236 CPU @ 3.40GHz
CPU cores: 1
Frequency: 3408.008 MHz
RAM: 217.0M
Swap: -
Kernel: Linux 6.18.32-1-virt x86_64

Kansas City, MO USA VPS named "kc":

~~~

nench.sh v2019.07.20 -- https://git.io/nench.sh

benchmark timestamp: 2026-05-20 21:28:57 UTC

Processor: AMD EPYC 7B13 64-Core Processor
CPU cores: 1
Frequency: 2250.000 MHz
RAM: 217.0M
Swap: 64.0M
Kernel: Linux 6.18.31-0-virt x86_64

The Vint Hill VPS is not running a swapfile, while KC is. Wouldn't this affect the performance benchmarks?

Not_Oles

@TemporiousOne said: The Vint Hill VPS is not running a swapfile, while KC is. Wouldn't this affect the performance benchmarks?

Yes, there is no swap file yet on friendly. I doubt it makes a difference. Tomorrow I can try adding the swap file on friendly or disabling the swap file on kc, or both, and rerunning nench. We will see.

Not_Oles

@Not_Oles said: Today's plan is possibly to try installing NetBSD again with these iPXE commands:

set net0/ip <YOUR_STATIC_IP>
set net0/netmask <YOUR_SUBNET_MASK>
set net0/gateway <YOUR_GATEWAY_IP>
set dns 74.82.42.42
kernel https://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/amd64/binary/kernel/netbsd-INSTALL.gz
boot

Ah, okay, this didn't work.

The iPXE shell said, after changing https to http,

iPXE> kernel http://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/amd64/binary/kernel/netbsd-INSTALL.gz
http://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/amd64/binary/kernel/netbsd-INSTALL.gz. . .  ok
iPXE> boot
Could not boot: Exec format error (https://ipxe.org/2e008001)

I guess the netbsd-INSTALL.gz image is not compatible with iPXE. Maybe that's why NetBSD is not listed on the BSD install menu mentioned above.

I ran the imgstat command suggested on the linked error page.

iPXE> imgstat
version.ipxe : 37 bytes [script]
netbsd-INSTALL.gz : 9554743 bytes [gzip] [SELECTED]
iPXE>

Not_Oles

@Not_Oles said:

@TemporiousOne said: The Vint Hill VPS is not running a swapfile, while KC is. Wouldn't this affect the performance benchmarks?

Yes, there is no swap file yet on friendly. I doubt it makes a difference. Tomorrow I can try adding the swap file on friendly or disabling the swap file on kc, or both, and rerunning nench. We will see.

@TemporiousOne

Here is the swap being setup on friendly:

friendly:~# date
Thu May 21 03:51:36 UTC 2026
friendly:~# swapon
friendly:~# ls -l /swapfile
ls: cannot access '/swapfile': No such file or directory
friendly:~# dd if=/dev/zero of=/swapfile bs=1M count=64
64+0 records in
64+0 records out
67108864 bytes (67 MB, 64 MiB) copied, 0.19616 s, 342 MB/s
friendly:~# chmod 0600 /swapfile
friendly:~# mkswap /swapfile
Setting up swapspace version 1, size = 64 MiB (67104768 bytes)
no label, UUID=c4a04e77-76a4-44ee-a3df-bd3727992a31
friendly:~# swapon /swapfile
friendly:~# cat /etc/fstab
# <fs>      <mountpoint>   <type>   <opts>              <dump/pass>
LABEL=/     /              ext4     defaults,noatime    1 1
friendly:~# echo "/swapfile none swap sw 0 0" >> /etc/fstab
friendly:~# cat /etc/fstab
# <fs>      <mountpoint>   <type>   <opts>              <dump/pass>
LABEL=/     /              ext4     defaults,noatime    1 1
/swapfile none swap sw 0 0
friendly:~# 

And here is another nench, on friendly, but now with swap.

-------------------------------------------------
 nench.sh v2019.07.20 -- https://git.io/nench.sh
 benchmark timestamp:    2026-05-21 03:59:23 UTC
-------------------------------------------------

Processor:    Intel(R) Xeon(R) E-2236 CPU @ 3.40GHz
CPU cores:    1
Frequency:    3408.008 MHz
RAM:          217.0M
Swap:         63.9M
Kernel:       Linux 6.18.32-1-virt x86_64

Disks:
vda      1G  HDD

CPU: SHA256-hashing 500 MB
    1.973 seconds
CPU: bzip2-compressing 500 MB
    8.690 seconds
CPU: AES-encrypting 500 MB
    1.144 seconds

ioping: seek rate
    min/avg/max/mdev = 169.5 us / 3.99 ms / 23.1 ms / 2.33 ms
ioping: sequential read speed
    generated 1.25 k requests in 5.00 s, 312.8 MiB, 250 iops, 62.5 MiB/s

dd: sequential write speed
    1st run:    300.41 MiB/s
    2nd run:    231.74 MiB/s
    3rd run:    313.76 MiB/s
    average:    281.97 MiB/s

IPv4 speedtests
    your IPv4:    147.135.xxx.xxx

    Cachefly CDN:         30.88 MiB/s
    Hetzner ASH (US):     29.98 MiB/s
    Leaseweb (NL):        0.03 MiB/s
    Online.net (FR):      9.13 MiB/s
    OVH BHS (CA):         32.92 MiB/s

IPv6 speedtests
    your IPv6:    2a11:6c7:xxxx:xxxx

    Hetzner ASH (US):     11.50 MiB/s
    Leaseweb (NL):        0.05 MiB/s
    Online.net (FR):      8.66 MiB/s
    OVH BHS (CA):         10.31 MiB/s
-------------------------------------------------

friendly:~$ 

Another nench on kc.

-------------------------------------------------
 nench.sh v2019.07.20 -- https://git.io/nench.sh
 benchmark timestamp:    2026-05-21 04:08:07 UTC
-------------------------------------------------

Processor:    AMD EPYC 7B13 64-Core Processor
CPU cores:    1
Frequency:    2250.000 MHz
RAM:          217.0M
Swap:         64.0M
Kernel:       Linux 6.18.31-0-virt x86_64

Disks:
vda      1G  HDD

CPU: SHA256-hashing 500 MB
    6.037 seconds
CPU: bzip2-compressing 500 MB
    35.095 seconds
CPU: AES-encrypting 500 MB
    10.262 seconds

ioping: seek rate
    min/avg/max/mdev = 384.5 us / 3.93 ms / 66.8 ms / 4.89 ms
ioping: sequential read speed
    generated 1.44 k requests in 5.00 s, 360 MiB, 287 iops, 72.0 MiB/s

dd: sequential write speed
    1st run:    17.07 MiB/s
    2nd run:    59.22 MiB/s
    3rd run:    78.58 MiB/s
    average:    51.63 MiB/s

IPv4 speedtests
    your IPv4:    208.84.xxx.xxx

    Cachefly CDN:         4.85 MiB/s
    Hetzner ASH (US):     5.38 MiB/s
    Leaseweb (NL):        0.02 MiB/s
    Online.net (FR):      4.04 MiB/s
    OVH BHS (CA):         3.82 MiB/s

IPv6 speedtests
    your IPv6:    2a11:6c7:xxxx:xxxx

    Hetzner ASH (US):     3.40 MiB/s
    Leaseweb (NL):        0.03 MiB/s
    Online.net (FR):      4.08 MiB/s
    OVH BHS (CA):         1.55 MiB/s
-------------------------------------------------

kc:~$ 

AnthonySmith

Another outage, hopefully they dont come in 3's

We lost Singapore for a few hours, and we're almost sure the cause has been found. Someone managed to get the hypervisor to do so much work that it literally murdered itself trying to keep up with the rate limiting, likely not deliberate, but it exposed a hole, so we have patched that now.

We also took the opportunity to intelligently batch some of the recovery options so in future if we have an outage the recovery will be much faster and more controlled, we also found people were allowed to attempt to migrate a VPS from a down hypervisor and attempt to delete a vps from a down hypervisor, not ideal, fixed.

Singapore needs more resources. Generally, it's WAY more popular than expected. I will get on that asap.

In the meantime, we have added a 25% max speed bump offset for Singapore and Australia networking, new VPS get this automatically; existing ones will need to upgrade and downgrade (any order) their network priority/tier on the VPS management page to get the new speed.