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
Nice waffles 👀
There is a bizarre issue with 2 dragon servers on-going.
They are RAID10, large arrays.
Stats show tremendous throughput and activity on drives and network, nothing EXCEPT testing with dd with iflag / oflag direct shows that somethings wonky, under 10MiB/s throughput.
Yet iotop may show 1.5-2GB/s at the same time, network might be doing few Gbps both ways.
Just haven't had time to look into this. might be some kernel limit that after certain number of I/O requests some just get slow. Tried swapping schedulers no use. Might be FS to array misalignment too -- or array is misaligned to drive sectors. Would've thought we would have caught that when the first server was built but you never know.
also since there is so many new users and servers right now, some servers can feel crawling -- everyone is downloading at maximum speed and settling in. These first couple weeks are always brutal.
Almost impossible to avoid tho, it's impossible to judge user prior to provisioning how heavy this user is. Just got to do best to load balance and hope luck is not against and many heavy users on same server. BUT with this many new users that's nigh impossible. I am sure @deank would say something like the end is nigh ??
I got these too. But more over.
I can't access to my seedbox anyways ssh, http, ftp and webui show #"504 Gateway Time-out"#
I create ticket and have to waited more than 12 hours. Until an hour ago I just got answered form the support that they reboot server. Now I can access to my seedbox and everything look nice as its should be.
Just refuse all clients. Easy as a halfeaten pie.
It is same. They replied my ticket so probably fixed my issue too. Will check it later.
At this price, I don’t expect much. I happy overall.
forgot to mention that you can always request reprovisioning if you do not need data migration. reprovisioning == fast.
Data migration == slow and takes effort, not very keen on doing any of those on per user basis.
@PulsedMedia
may I ask a question? the upload has always 100Mbps limitation even when the traffic limitation not exceed? wget the public files in the box seems been limited to 100Mbps, even between the boxes in the same datacenter, wget also can always only get max 7~8MB/s
thanks
@PulsedMedia someone on my Astro's 12TB node is hitting hard on the disk for around a day. Disk read is much larger than network out, so maybe they are verifying the files. My shell is really unresponsive right now. Is that kind of usage considered unfair or abusive?
What? It means not having such a large sale or giveaway to overwhelm the servers. Simply staggering new users and round robin adding servers to spread out load. But instead, you use unused servers for mining or chia or something and bring additional servers online when a minimum user load would be reached. This was discussed a previous time when people brought up the crypto stuff on "unused" cycles.
Yep thats limited intentional by .lighttpd.conf in your home, modifiying it will trigger re-creation of the defaulft file and putting another limit into /.lighttpd/custom will not overwrite the value
sftp on the other hand is kinda cpu heavy, so the best speed i only get out of old unsecure ftp transfer
Same here, read speed is kinda okay but any write to the disk barely reaches some MB/s - was tinkering with torrent settings for quite some time now... When downloading it always starts super fast >100MB/s but stalls after seconds - until i found out now it's caused by disk write speed
basically Initial speed is good until cache is full and it has to write down to disks
Well then, the 100Mbps of lighthttpd was intentionally restricted? if no, is there any way can change it? thanks.
our provisioning algo goes much beyond simple round robin, unfortunately we were unable to code an crystal ball to the provisioning algo to guess how high load each user generates on their first few weeks. It does try it's best tho and servers with new users are "penalized" in priority.
Do you know where we can get such a crystal ball? I heard there's a shop which sells them, the price being wholesale box of burnouts. Kinda steep if you ask me.
Or perhaps you meant we should put everyone on a separate dedicated server for the first 2 weeks? Just keep 500 servers laying around so everyone gets a dedicated? Let us know how we can do that, we would be happy to put over 10x the capacity of servers idling if it has no financial drawbacks.
Another presumption you made is that we only add servers after "minimal" load. Before this celebration began there was quite a few servers with 0 users AND multiple times more vacant capacity than the initial 4x 100 accounts combine to. Thing is, adding only when absolutely necessary does not work when you have your own DC, you plan 6-12months ahead. If you use 3rd party DC dedi servers, sure then just add when needed to, but when you own the DC you have to plan way ahead.
Btw, chia is not mined. It is harvested. Harvesting takes almost zero resources, people have petabytes running on a RPi with USB Externals, hub after hub and still the RPi mostly idles. So yeah i guess it's kinda heavy and ruins servers.
Using the service for normal tasks is not abusive. You are on a shared server and thus sharing the resources with other people, it is normal to have fluctuation.
Besides those numbers are far from typical 12 drive server max performance, and actually kinda wrong metrics too (I/O latency is more important than throughput). MEGA special mostly or solely went to 12 drive servers.
can we use these seedboxes as backup servers ? like can we download/upload data from ftp, rsync, etc ? can we install programs using ssh ?
Yes you can use for backups, and you can install what you can within your userspace.
Yes, sftp, ftp, rsync, scp, lftp etc.
Thanks for the clarification. So, its basically a vps but why is it especially categorized as a seedbox? maybe because of preinstalled software or a web UI you have developed to help with seeding ?
it's not a full VPS and specialized for seedboxes, not regular VPS.
HAPPY 12th BIRTHDAY!
remember they are raid0(m10g is raid10?),use with caution.
Given that installing applications in user space is allowed, you can install another HTTP server that doesn't have artificial speed limitations.
However, nginx on port 443 would not reverse proxy to this HTTP server.
You would have to acquire a certificate and do TLS termination yourself, on a non-standard port.
alternatively there is a filemanager-share rutorrent plugin to directly share your downloads via http, as this runs over rtorrent ports it should not be limited
https://github.com/nelu/rutorrent-filemanager
edit
most probably also limited as it runs over same lighttpd rutorrent instance
My account is on a 4-drive server, so that seems to be the maximum. Should I request a re-provisioning then? My 12TB box used to be usable, but now it's very unresponsive
EDIT: I/O wait seems high too
@PulsedMedia i just got purchased 12th Anniversary V10G XS but upload speed far slower than my current plan with 1Gbps network
12th Anniversary M10G 1TB stock
I am still hoping for a Very Fat32.
@PulsedMedia
wget the files on Hetzner Germany are too slow, only several MB/s, even often under 1MB/s, both Nuremberg and Falkenstein are all have this issue, but Hetzner Finland is very good, could you please help check this issue? perhaps some route issue? I have asked some friends, their box are all have the same issue.
thanks.
above comments pretty much confirms, pulsed oversells.
I might as well pay for 3+ month, and use after 2 months
when demand die down
Both direction MTRs are required.
Regardless, this is most likely Hetzner side routing, it's their network. We have fast connectivity to their Finland DC, why don't they route their traffic between datacenters internally and announce all their networks at both ends?
Probably because long distance point to point (no matter CWDM, DWDM, Dark fiber) is super expensive, and it's cheaper to go via transit.
They probably only have direct links between datacenters used for their own servers, but do not announce say Falkenstein IP ranges in Finland, and vice versa.
It quickly looks like 3rd party issue, where in reality Hetzner is just trying to save money.
We cannot fix how they announce their IP ranges, and traffic will route by whatever transit is the shortest length.
Get both ways MTR from all 3 locations and we can see if my suspicion is confirmed.
Then see MTR between all 3 of the servers.
Very interested in seeing those results!
Does anyone who has Astro's MEGA plan see above 10mbps?
@PulsedMedia How do you divide port speed per user? 10gbps port speed but I can't go above 10mbps.