BERO HOST 10th Anniversary Deals – EPYC Turin / Ryzen 9950X / IPv6-only VPS

berohost

@cnblackhole said:

@berohost said:

@hellomiller said:
when the 10th-anniversary limited edition model will be restocked?

I am planning to top up my account balance right now, but I have a concern: if the item sells out again due to high demand right after the restock and I fail to secure one, will I be able to request a refund for the topped-up amount at that time?

Restock in 2-3 days.
If you don't receive a server during the restock, please contact us again via a support ticket afterward. If there are no resources left, we will, of course, refund your money.

Hi, one more question — if I top up via PayPal but don't manage to get a VPS during the restock, would I receive the full amount back, or could there be any fees deducted from the refund, whether from your side, PayPal, the bank, etc.?

If you do not receive a server, please contact us again. If no resources are available, you will receive a full refund.

berohost

@abitstack said:
Oh man, I just double checked my shiny new machines on the anniversary cluster after some days and I am still hesitant to put any production workloads on them ...

While everybody is yapping at geekbench scores those cpus are plenty fast and not oversubscribed. But what matters for prod - disk and net - are toast. Nvme on all of my machines is depleted down by roughly 30% and worse totally unpredictable, so none of them is the right one for the DB server.

Incoming new servers might help spread out the load if the next batch does not get immediately oversubscribed just like the first one. (and yes, I get it that this is a business and not a pet project, but still;)

Anyway, more hardware is nice and might help mitigate the load on the hosts
BUT the real elephant is the network! - its smoked
However I plan to combine the servers there are rtt (ping) spikes between them all the time and timings go up and down unpredictably. Traffic between the VMs gets sometimes > 1k drops/retries per second since the switches seem not to be able to cope and buffers get overrun. Extra (core) switches are not cheap but can help a lot. Right now the net is fried -kfc.

Are there some plans to extend/segment the network further with all the new machines coming online ?

There are no known network issues, and we monitor our switches very closely. Please feel free to contact us via a support ticket so we can look into it.

abitstack

@network said:

@abitstack said: disk and net - are toast

How slow is "toast"? YABS?

I did not take screen shots but the servers started at ~140k 4k iops and are all down in the 90es right now - in comparison the old machines are still at +/- 130k. This is a whole new neighborhood:).

The problem is I am not running a file hoster to max out those 25TB - I am fine with a couple of GBs (using a CDN for static assets) - but I need predictable response times for the APIs.

This boils down to consistent RTTs between the servers and stable response times from the nvmes for the DBs

If you have a machine check a tcpdump then you know:)

meowwcc

Man, my VPS network just abruptly went down just now... Unreachable via ping, ssh etc

Request timeout for icmp_seq 0
Request timeout for icmp_seq 1
Request timeout for icmp_seq 2

On my hosted website (cloudflare DNS) it also reports Host Error.

Restarting the VPS from their panel didn't resolve it. Not sure what is going on but it's definitely a network issue on their end.
Created a ticket in the mean time

vitobotta

Who else's servers are unreachable?

vitobotta

I don't know if i am just unlucky but people here were saying great things about uptime, and here I am, just a few days in, with an outage already.

nick_

Yup, down for me.

Hitori0221

service down, luckily I have hourly backup

vitobotta

I am starting to regret migrating from the dedi I had already.

Lkwang

This news really shocked me. SLAs are something a lot of people care about, right?

network

All my servers are down as well. This is the first downtime since Black Friday though.

network

VNC is showing:

Failed to create new system journal: No space left on device.
systemd-journald []: Failed to create new system journal:
No space left on device

What does this mean?

network

All are up now.

haolun

All my servers are down as well.

vitobotta

Mine too. Hope we can get an explanation soon.

meowwcc

@network said:
All are up now.

Still down for me. Not accepting ssh, websites aren't up

network

The uptime is same as before, so there was no reboot. Network problems?

abitstack

@berohost said:

@abitstack said:
Oh man, I just double checked my shiny new machines on the anniversary cluster after some days and I am still hesitant to put any production workloads on them ...

While everybody is yapping at geekbench scores those cpus are plenty fast and not oversubscribed. But what matters for prod - disk and net - are toast. Nvme on all of my machines is depleted down by roughly 30% and worse totally unpredictable, so none of them is the right one for the DB server.

Incoming new servers might help spread out the load if the next batch does not get immediately oversubscribed just like the first one. (and yes, I get it that this is a business and not a pet project, but still;)

Anyway, more hardware is nice and might help mitigate the load on the hosts
BUT the real elephant is the network! - its smoked
However I plan to combine the servers there are rtt (ping) spikes between them all the time and timings go up and down unpredictably. Traffic between the VMs gets sometimes > 1k drops/retries per second since the switches seem not to be able to cope and buffers get overrun. Extra (core) switches are not cheap but can help a lot. Right now the net is fried -kfc.

Are there some plans to extend/segment the network further with all the new machines coming online ?

There are no known network issues, and we monitor our switches very closely. Please feel free to contact us via a support ticket so we can look into it.

Ok, thanks for the swift reply as always !
As stated before I am probably better off being a bit more stoic and wait until the frenzy calms down. Will wait until the haywire of the next batch is over, then we can sort things out.
I am more interested in the long term outcome - just not very patient (mea culpa).

As for the net - throughput is fine, but pings spike and base latency fluctuates inside the cluster (bad for DB & API response time predictability).
There are scan storms on the network right now - they don't hurt the throughput since retries are cheap but overwhelm the switches with small packets until they start dropping and pings spike. Thus the question for segmentation/more switches to tame the packet storms.

meowwcc

ssh works now but throughput is still terrible, damn

Falzo

@vitobotta said:
Mine too. Hope we can get an explanation soon.

for what? a blip in the network, that might not even have been in their DC? did you do an mtr?

vitobotta

@Falzo said:

@vitobotta said:
Mine too. Hope we can get an explanation soon.

for what? a blip in the network, that might not even have been in their DC? did you do an mtr?

What is wrong with you? Is it surprising that I would like an explanation from the provider as to what caused an outage? Some people in this community are truly unbelievable.

Falzo

@vitobotta said:

@Falzo said:

@vitobotta said:
Mine too. Hope we can get an explanation soon.

for what? a blip in the network, that might not even have been in their DC? did you do an mtr?

What is wrong with you? Is it surprising that I would like an explanation from the provider as to what caused an outage? Some people in this community are truly unbelievable.

agreed :-)

abitstack

Thumbs up for resolving the network outage that quickly (<30 min) !
Unfortunately still ARP+SSDP+solicitation packet floods and fluctuating internal pings, however good job

taizi

i sent a ticket to upgrade my server yesterday,but seems no response yet

hellomiller

@berohost said:

@hellomiller said:
when the 10th-anniversary limited edition model will be restocked?

I am planning to top up my account balance right now, but I have a concern: if the item sells out again due to high demand right after the restock and I fail to secure one, will I be able to request a refund for the topped-up amount at that time?

Restock in 2-3 days.
If you don't receive a server during the restock, please contact us again via a support ticket afterward. If there are no resources left, we will, of course, refund your money.

Thank you for your prompt and reassuring reply. I have topped up my account.

berohost

There was a brief network issue that lasted about 15 minutes. We are currently investigating the exact cause and will provide an update once we have more information. Since then, however, everything has been running smoothly again.

amsaal

@berohost said:
There was a brief network issue that lasted about 15 minutes. We are currently investigating the exact cause and will provide an update once we have more information. Since then, however, everything has been running smoothly again.

could you please provide feedback on ticket , appreciate i want to upgrade from Aniversary S to M .. i understand your busy with support.

berohost

@amsaal said:

@berohost said:
There was a brief network issue that lasted about 15 minutes. We are currently investigating the exact cause and will provide an update once we have more information. Since then, however, everything has been running smoothly again.

could you please provide feedback on ticket , appreciate i want to upgrade from Aniversary S to M .. i understand your busy with support.

We will reply to you this evening

vr10

Three years in and still going strong. I was tempted to buy another, but we'll hopefully get a top up bonus for Black Friday

vitobotta

Is it possible to transfer these servers to other customers?