It’s just for me or anybody else notice the sudden spike of network/disk IO from the last week? It’s properly 3x-4x of normal amount of activity. Some of my old disks can’t handle the IO anymore and my router will die soon with this kind of traffic. Lol. I’m using storagenode version 1.91.2.
Low traffic - we (SNO) are complain “why so low?!”, a lot of traffic - hm, we complain again - “our routers dying in fire!”
No ideal balance I would say
If serious - you need to provide an geographic region also.
Because my nodes feel OK
This is inevitable. Being able to run node on a naked rotational hard drive was a temporary thing. You only have few hundreds iops bandwidth at best.
You can get a datacenter class SSD on eBay for $10. Convert your single disk to ZFS pool and add the SSD as a special device. This will get you a lot of iops headroom.
Replace your router. It’s trash.
“use what You have” ?
yeah but mostly what ordinary people have for personal use is trash and can’t handle huge amount of traffic. So… By the way nobody notice any huge increasing of traffic?
US location, normally was 20-30GB ingress per day before, now 50-60GB ingress and disk/CPU constantly working. I think something changed in latest storagenode version but not sure.
This is the typical “mining” mindset. Storagenode versions don’t give you x% increase in “stor(j)ing”
I don’t understand what you talking about. I’m not talking about any “storjng”. I’m saying that maybe some changes in filewalker or sth in new storagenode version may lead to excessive disk/CPU using. I’m too lazy to read/understand the changelog. That’s why I’m here for you experts to explain to me. Lol.
Okay so nobody notice any strange spikes? Fine. I’ll keep monitoring the situation. Thank you.
But if what you have is not up to task — then you either stop hosting node, or make improvements.
Like paying $10 bucks to get an SSD, which will also accelerate all your other usecases.
Something tells me running storagenode on a single HDD is already violating the “use what you have” rule — if not for the storagenode that HDD would not be spinning. There is no good usecase where running single hdd 24:7 is justified.
It’s not only about quantity but quality.
Upgrading the router is the single most impactful thing one can do to improve their networking experience. Regardless of whether there is a storagenode in the mix. Most people are in awe at what they were missing all this time.
Things like jitter, latency, and bufferbloat are affecting modern applications more than ever — video and audio conferencing and gaming, especially, but also mundane things like browsing and downloads.
If we include wifi access point into the mix — the abyss between “most” and “good” are vast.
You want robust SQM, solid modern wifi — at least 801.11ax, _ honestly implemented_. You don’t want mesh. You want a lot of ram and fast CPU (or hardware offloader of some sorts) to manage all that. None of that exists on cheap devices.
Getting a good one does not need to be expensive. There are plenty of nuc-type single board Chinese computers with two or more ports, they are under $50. Add free firewall like OpenWRT or pfSense — and off you go. Instant improvement. Turn your existing router to an access point (most suppprt bridge mode). Then replace it with an actual enterprise access port, budget permitting (around $100-200).
There is no excuse to tolerate tplinks and Synologys and netgears and asuses. Albeit asus is one of the better ones — it’s still consumer garbage.
Yea i know, just pointing absurdity of the mantra. and whole approach of STORJ inc. to a SNOs.
Just prooves my belives, that STORJ need more profesional SNOs, who should be treaned more like workers, than occasioanl opportunnist to park “unnused HDDs” for behalf of STORJ.
I fully agree. A regular home user cannot run and maintain a node. But they are not a problem – home users generally don’t have hardware to host it to begin with, skills to configure it, or desire to do so in the first place.
The gap is when the user still does not host any services, and does not have any hardware aside of old scraps, but is tech-y enough to understand how to run the node. Then suddenly, when scaling becomes an issue, all falls apart.
I think the message should be that storagenode is intended to be utilizing spare capacity, not _run on an ancient abandoned obsolete pieces of tech laying around the garage because one did not bother to take to the recycler". I.e. “don’t buy anything” is not the same as “use garbage nobody else wants”
I can confirm, traffic did go up:
“Don’t buy anything” means “don’t buy anything” so you should use what you have. Especially since the traffic was so low for months now (for a new node probably even lower) that the old unused hardware could handle it.
Also, remember, for a long time, the recommended way to run a node was a Raspberry Pi with a single hard drive (and no SSD). You can’t fault people for following recommendations.
Really!? I wasn’t aware of that. It’s a horrible advice. 8GB raspberry pi with carefully configured SSD caching may actually work. But I digress.
Nitpick: If that was unused hardware – then by using it you are technically buying electricity for it.
On topics, I see small increase in traffic too:
Not sure if that site reliable, but You can check and see it shows it too,
not sure if new forum topic necessary every time it spikes.
You can see some cool increase in ingress especially:
not sure if entirely reliable,
the site shows also 3 segments lost lastly on production, and was 0, always.
It would be interesting to see a “heat map” of all storage nodes in terms of egress/ingress over time, how the network “breathes”. Aggregated numbers don’t tell much.
+1 oh wow,
We need a “Super Heart” on this forum, like a Golden button in talent shows,
once every 6 months, one could give such to really highlight some valuable post! Lol.
Likely customers are more active in your region, could you please check - what’s traffic doing the spike?
Hover a mouse above spike on the bandwidth usage graph and take a screenshot and post it here.
I have Raspberry Pi3b+, 1GB RAM, 2TB 3.5" ext4 via SATA extension board, in the metal case with the fan, it was running since 2018 for 4.5 years without huge issues (the OS is on the SD card and I reflashed it 3 times, but when it was require doing so in 4th time, I was unable to return and fix it, so it is stopped to boot and I switched it off via smart plug).
This node worked pretty well, it accepted less traffic than my Windows node on i7 home server, but not so far (about 10%-15% less), so I would say it was pretty capable even during stress tests which have happened in the beginning of the network, when I have had load of 80Mbps in both ways with spikes up to full bandwidth of my network (100Mbps). It was the only node that time.
One of the options you have if the current node can’t handle the traffic would be to add a second node so the traffic is split across both. Of course you would be paying electricity for a second drive then
I virtualised my routers over a year ago now and they have worked very, very well. It also makes recovery easy since if the current hypervisor fails I can just restore the vm to another system, plug in the network cables, boot and we are back. I’ve configured my three hypervisors almost identically so restoration is easy as well.
Yes, we are using tp link AP’s in bridge mode - no need to improve them though as they are just to keep the kids computers on their own VLAN. That VLAN cannot access the main home network.
I’m just waiting for the kids to get ransomed at some point. lol There is also an IoT device in that VLAN as well that I also don’t trust. (Chinese Android TV device)