in regard to the router, do something so it will get rebooted every day or week… they should be rather stable but sometimes they will have like memory leaks or something in their programming which can lead to disruptions.
any sort of timer might do fine, set it to turn off once a week for like 15minutes if its an old one without to many options, ofc i suppose it might be a 24hour one… but still we are allowed like 288hours so you might be able to have it shutdown for an hour a day and still have 200 hours to spare every month.
so atleast as a temporary fix to test if it will work should be fine, tho your online score would drop ofc.
else i run my setup with ecc and redundancy so most issues can wait until i can be bothered fixing it.
have this hdd that is running terribly clicking away… due to the cold in my server room, but it seems to run fine… causes a slight bit of latency but nothing huge, so i just leave it there because i don’t have anything to replace it, and really it’s been doing this for months when it gets cold…
and because i have redundancy i can allow to take the risk to run it into the ground… ofc costs twice the space when one is running mirrors, but in return i get much more control over my hours.
and with our ingress as it is at the moment… having nodes die… well i don’t really consider it an option… so looking at the ingress this month then its 10gb avg pr day for my main node.
so multiplying that by 1000 aka adding 3 zeros gives us 10TB in 1000 days of ingress.
1k days is about 3 years, so unless if you are running enterprise hdd’s then thats even beyond the hdd warranty, mean the odds that it will die is pretty high.
if it dies you loose the held amount ofc thats only like half, and you got like a fair number of months in with profits… but still 1 year in and you will have just started getting 100% payouts, 3 months from getting your 50% held back.
but still you will have like 3.3 tb data so it’s like 15$ a month on a hdd that would cost 10x that…
so the first year is basically just paying off the hdd, 10tb lets say…
maybe it’s more 15$ seems kinda low here it’s like 25$ pr tb… but lets be optimistic and say a year then. we can take the 2nd year’s profit and put towards hardware since the hdd will be the main part of that. and maybe we can even say thats also the power costs.
the first year peak is 15$ pr month, but realisticly thats more like lets say 90$ total
earned in year one, so year 3 would be at the 6.6tb mark… going on 10tb and full payout so thats the real earner.
so at 4$ pr tb that is lets say 25$ pr month for 6.6tb so 300$ for 12 months + 90$ surplus for the first year.
so looking at that math, having to put hourly wage of keeping it running for 3 years, internet costs (even tho the internet requirement is rather low.
so what can we learn from this… well lets add another year…
year 4 of a full 10tb at 4$ pr tb so 40$ a month so 480$ earned
so really if a node dies, in say the first 5 years its earning potential drops like a stone, and really the longer nodes can run the better the economics of it becomes.
i know people see it as wasteful to run with redundancy and they hate mirrors even more…
one could ofc do a raidz1 but then quickly the space saved increases the power costs.
which for me is rather important because i have very expensive power.
so having like 5 hdd’s in a raidz1 basically increases the costs of the power that what i use for power i could pay for the mirror disk instead.
and mirrors rule… lol
this is getting a bit longer than expected… but i have a point to make…
and it is this…
from my perspective there is absolutely no benefit to running without redundancy and using ecc memory.
the cost vs the added work(from bit flips and write errors), the erratic hours, decreased avg life span.
it just isn’t worth it long term… sure one can run single disk nodes in the beginning without to much risk, adding a mirror later isn’t to much work… but still errors that are caused then can show up and cause issues at later days… so i run mirrors all the way.
ofc i may try to run some nodes on single disks and without ecc just for the experiment to see if my evaluation is correct… ofc getting data back on something like that could take years…
so yeah… get some redundancy so you can sleep at night … 