New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
today
classic story: it slowly becomes worse and worse and will probably be unusable in a few months. the server will expire soon so unfortunately i won't be able to post updates!
Ralph should stick to selling books. He's always been rude and his services are thrash.
What are you using to monitor your servers? (in particular the first and last screenshots)
I'd like to do the same.
Uptime kuma is first one, ping.pe is the second
It appears HostSlim's entire 69.174.98.0/23 range is unreachable at the moment.
Probing 69.174.98.1 shows it belongs to RIPE NCC, AS3257 (GTT-BACKBONE), GTT – Estonia, Harjumaa, Tallinn [hosting]. ( hmm, I have seen this shit before, literally like 24 hours ago )
The prefix 69.174.98.0/23 is clearly part of a larger aggregate, 69.174.0.0/17.
Unless HostSlim has actually withdrawn the entire /23 — which I highly doubt — this smells like pure GTT fuckery.
We’ve just gone through the exact same bullshit recently: a /24 we rented out suddenly started being announced again as part of the larger /19 it belonged to. After warning the broker for over two months that the subnet was being split incorrectly, I eventually dropped the entire subnet and told them to shove it up their ass.
Their reply? “Yes, but the more specific announcement takes precedence.”
Like I don’t know how BGP works.
I may not know everything about BGP, but this is basic stuff: if you carve a larger block into smaller prefixes, you do it properly. You withdraw the aggregate and announce only the /24 ( the active subnet depending on the split ). You don’t leave the larger prefix active and hope the Internet magically behaves.
Announce the /24 and be done with it. Anything else is sloppy at best and operationally broken at worst.
The subnet in question of HostSlim looks like the shit we had:
I personally doubt this is on them ( HostSlim ), I would say it is on the holder of the larger prefix ( this case GTT )
EDIT:
I’ve been seeing this more and more often lately. It genuinely feels like people are getting dumber at the NOC level.
I understand aggregation. This is not an aggregation problem.
We’re talking about larger prefixes being actively rented out as smaller blocks, while the parent aggregate is still being announced. That’s not a design choice, and it’s not a policy debate — it’s a lack of basic operational skill.
If you lease out a /24 from a /19 or /17, you either withdraw the aggregate or you very clearly control how it’s announced. Leaving the larger prefix active and relying on “longest prefix wins” is lazy, error-prone, and eventually causes exactly these conflicts.
This isn’t BGP being complex. This is people at the NOC not understanding — or not caring about — the consequences of what they announce.
The Internet still works because of math — not because of people doing their jobs well.
i was wondering why smokeping turned green since yesterday, lol (my ip is in another subnet which works)
im returning once again. the network is still shit. ssh drops every 5-15 minutes, udp is very strange: quic works, but not openvpn. it was very strange in the last few weeks: multiple times i noticed that when openvpn worked, later it would break and only small packets would go through the tunnel. idk if this is an anti-ddos filtering or what, there was no notification from hostslim, as always
Zero issues whatsoever with the Estonia location over the past months or right now, Amnezia WG works flawlessly and there is no packet loss or ssh drops.
Maybe it's your ISP or your country's censorship?
connect to ssh directly and wait.
a few days ago i was connected to it and to another server, both over a VPN (which is on another, third vps), hostslim kept dropping the connection. today i connected to hostslim directly and it was still dropping my ssh connection.
Sure, left it on for now.
If you have NAT on your ISP, that may also affect things. Check out
ServerAliveIntervalin/etc/ssh/ssh_config(on the client).That's indeed correct, a long-running SSH connection (at least with nothing launched) hangs after a while. Typing no longer works, but there is no immediate disconnect either.
ServerAliveIntervaldoes not help, the connection will then end instead, with a "Timed out contacting server" error message.But then I SSHed into my server inside WG, using the in-WG IPs. And that one did not hang even after a long time. So it appears this is caused not by any dropouts in the network, not by any packet-loss or the like, but rather a misconfigured firewall on their side or their upstream.
it hangs even while typing and reconnect is required, also ssh port doesn't matter
ssh stopped dropping and openvpn started working. im like 90% sure it will break again
This network sounds awful in a beautiful way. Thank you for sharing this information with the world. I will be buying a server or two
@TrikeLike i saw your topic and you probably won't find it useful imo. last year it was like "netem random 10% loss", a few days ago it was like "iptables connbytes 1000 drop", currently it's fine and you will never know when something will happen.
honestly i haven't seen a network worse than "netem random 50% loss reorder 10% delay 100ms 50ms" on LAN, if your program survives that then it will be like 90% fine anywhere. (the best thing that i could find for such conditions was "https://github.com/apernet/tcp-brutal" tcp congestion protocol (btw their quic implementation was much worse in case of packet reordering or jitter), for others you may look at my sysctl.conf https://pastebin.com/raw/nixHq56s)
i mean, there are other cases such as "16kb block" and "blocking of multiple TLS sessions" in russia for example (you may read about it and some other cases at https://habr.com/ru/users/0ka/articles/), but i have no idea if you are affected by that
Sounds exciting. Do they charge extra for the experience would this be a little free thank you coming with every plan?
only icmp seems to work fine right now, any tcp and udp ports work for just 15 seconds every few minutes. also i forgot to add that my openvpn didn't work yesterday (idk since when)...
upd: since icmp seemed unaffected, i disabled icmp echo on my vps, and when the network broke again, suddenly my ip address started responding to pings, that probably means someone hijacked my ip address
someone is still jacking my ip address
(empty - good, line - someone else is using my IP)
whole /24 is down for 4 hours already
Unfortunately we need to be on lookout for this issue may have returned.
2 days ago immediately after the introduction of new IPs the network started jittering all over the place, and there is now packet loss. I posted my thoughts in the thread, no response from the provider.
Now the next hope is 5th June or whatever, when the old subnet IPs get removed. Maybe having 2x as much IPs on the node is overloading the system in some way.
The graph above is from IPv6, since there's no easy way to plot a contiguous graph for IPv4, as those got switched over. But it's showing that IPv6 got affected too, even though it wasn't changed in this case.
same for me, and now it's at 1% average packet loss in the last 24h, before it was at 0.2%