Consolidating everything or keeping things in different VPSes?

OujiOuji OG
edited December 2019 in Technical

Hi, because of the BF/CM I went on to think again about consolidating my services in a dedicated one.

Currently my dedi is a Hetzner with the following specs

code
i7-4770
32GB RAM
2x2TB HDD

Currently I have several servers, mostly KVM, but I few OpenVZ. I'm gonna list those in order:

2vCPU/4GB/110GB HDD/5TB
1vCPU/1GB/16GB NVME/10TB
1vCPU/512MB/250GBHDD/1TB
1vCPU/3GB/3TB HDD/16TB
1vCPU/1GB/15GB HDD/1TB - NAT OVZ
1vCPU/2GB/100GB HDD/1TB - OVZ
6vCPU/6GB/75GB SSD/5TB - Obviously OVZ

As you guys noticed, I'm kinda addicted in buying server (thanks to you)

Now I have a few unpaid invoices and I'm wondering about not paying them and maybe cancelling the other services I have, but I'm not sure, so I'm seeking for advice.

The invoices I've left are:
Avoro's BF special
Contabo's 8GB/200GB SSD
servarica's 500GB ipv6 only VPS

Currently I don't host a lot of stuff, most self hosted applications from the awesome-selfhosted github, but I'd like to be able to scale easily if needed (I've had requests in the past to host specific things).

Thanks for the input!

Consolidating everything or keeping things in different VPSes?
  1. Consolidate or nah?31 votes
    1. Consolidate it in your dedi, cancel everything else
      12.90%
    2. Consolidate it in your dedi, but keep or one two VPSes for service availability
      77.42%
    3. Cancel your dedi and keep the VPSes only
        9.68%
«1

Comments

  • Consolidate it into dedis, but keep the vps as is.

    • pro idler.
    Thanked by (1)uptime

    If it’s not broken, keep fixing it until it is. Blink twice if you agree.

  • WSSWSS OGRetired

    I'm a bit nuts with how I handle mine.

    Most of my important shit is external. I have 4 places across the world with 3 different primary transits for DNS. Mail is external, but sync'd to another location on RAID1 just in case- I haven't lost anything since 1998, and I don't plan to do so now.

    Everything forward-facing goes to a 10+ year old shitbox that costs about as much as a cola a day. Do I have backups of that? Sure. Do I care? Not really.

    Thanked by (3)MichaelCee uptime vpsgeek

    My pronouns are like/subscribe.

  • Cancel your dedi and keep the VPSes only.

    your dedi would be a single point of failure.

    The all seeing eye sees everything...

  • As per the low end spirit, you idle'em.

    Thanked by (2)Ouji vyas

    ♻ Amitz day is October 21.
    ♻ Join Nigh sect by adopting my avatar. Let us spread the joys of the end.

  • @somik said:
    Consolidate it into dedis, but keep the vps as is.

    • pro idler.

    While I like that idea, my country's currency is going downhill and I'm thinking about cutting costs.
    If I keep everything as it is, I'd be paying up to 60% of my monthly wage every year, which is a lot to me.
    And yes, most of them are just idling. I got them in the heat of the moment cause sometimes you feel like the deal is too good to pass.

    WSS said: I'm a bit nuts with how I handle mine.

    Most of my important shit is external. I have 4 places across the world with 3 different primary transits for DNS. Mail is external, but sync'd to another location on RAID1 just in case- I haven't lost anything since 1998, and I don't plan to do so now.

    Everything forward-facing goes to a 10+ year old shitbox that costs about as much as a cola a day. Do I have backups of that? Sure. Do I care? Not really.

    That's nice. I was thinking of keeping the dedi and a few of the VPSes (the two storage ones) for backups. And one or two to host my services in case of failure on my dedi.

    @terrorgen said:
    Cancel your dedi and keep the VPSes only.

    your dedi would be a single point of failure.

    Like I mentioned in this reply, I might keep the dedis and a few of them. I'm not sure yet. For now I don't have mission critical services and I could go on without them for a day or two, maybe a week.

    Also, it does feel so much easier to just install Proxmox on the dedi, 'cause backing up with Proxmox is so easier than one separate systems.

    Thanked by (1)vimalware
  • I find one large system allows for much easier scaling of services that end up requiring more resources then initially anticipated as you can just allocate more memory as needed and can make as many containers or virtual machines as you might want for no additional cost whilst being able to utilize 100% of the resources 100% of the time if what you are doing calls for it. I do have 1 NAT in another region just for things that benefit from being there that don't require much resources.

    Thanked by (1)vimalware
  • I would not suggest to put everything on one big machine.
    Because if this thing goes down, you gonna have a bad time.

    Better split stuff up.
    Even hole datacenters can loose power and/or network, we have seen such stuff multiple times.

    Even if you have 3 different power sources, you can still fuck it up, like OVH did.
    The hole DC went blyat.

    Thanked by (2)poisson vpsgeek
  • cybertechcybertech OGBenchmark King

    @Neoon said:
    I would not suggest to put everything on one big machine.
    Because if this thing goes down, you gonna have a bad time.

    Better split stuff up.
    Even hole datacenters can loose power and/or network, we have seen such stuff multiple times.

    Even if you have 3 different power sources, you can still fuck it up, like OVH did.
    The hole DC went blyat.

    We need to have redundant datacenters then. Someone's gotta build one beside OVH Leaseweb Hetzner

    I bench YABS 24/7/365 unless it's a leap year.

  • poissonpoisson OG
    edited December 2019

    @Ouji You clearly have an addiction, like we all do at some point. :)

    I think you should just work out what are the CPU, RAM and storage requirements and consolidate on a well-managed VPS offering optimum performance nearest to you (you are in China, right? I am in Southeast Asia and I find @seriesn's Nexus Bytes Germany location an excellent compromise for bandwidth speeds to Asia).

    For the sake of redundancy, I would find another VPS that is lower in price and specs as a failover making sure both are mirrored (location won't be an issue here since this is just in case shit happens).

    Finally, if you have a dirt cheap VPS storage like @cociu's offer, that would be the final backup you need.

    Dedicated is nice, but it is cheaper to go with VPS for redundancy. Single point of failure is not fun, even for self-hosted applications.

    EDIT: forgot to address the scaling part. I don't think you should be paying for things that are unlikely to happen. If it happens, I think just get another VPS from a good LE provider to meet the need. I mean if you are going to pay for a dedi for this not very likely to happen scenario, you can probably save two-thirds using a few VPSes and if the request comes in 6 months later, your savings will more than pay for a VPS to do the job, especially if you consistently work with a friendly provider who can probably do something to fit your needs.

    Thanked by (2)Ouji seriesn

    Deals and Reviews: LowEndBoxes Review | Avoid dodgy providers with The LEBRE Whitelist | Free hosting (with conditions): Evolution-Host, NanoKVM, FreeMach, ServedEZ | Get expert copyediting and copywriting help at The Write Flow

  • I try to size my proxmox kvm disks slightly under the common template sizes, so that I have a vps exit strategy. (I.E dd over the converted vma backups from my backup-vps to a new KVM VPS, and growfs if necessary.)

    Hetzner sb dedi can't be beat for RAM hungry apps ; Currently at 11-12GB utilization on idle.

    Thanked by (2)Ouji Hetzner_OL
  • @poisson said:
    @Ouji You clearly have an addiction, like we all do at some point. :)

    I think you should just work out what are the CPU, RAM and storage requirements and consolidate on a well-managed VPS offering optimum performance nearest to you (you are in China, right? I am in Southeast Asia and I find @seriesn's Nexus Bytes Germany location an excellent compromise for bandwidth speeds to Asia).

    For the sake of redundancy, I would find another VPS that is lower in price and specs as a failover making sure both are mirrored (location won't be an issue here since this is just in case shit happens).

    Finally, if you have a dirt cheap VPS storage like @cociu's offer, that would be the final backup you need.

    Dedicated is nice, but it is cheaper to go with VPS for redundancy. Single point of failure is not fun, even for self-hosted applications.

    EDIT: forgot to address the scaling part. I don't think you should be paying for things that are unlikely to happen. If it happens, I think just get another VPS from a good LE provider to meet the need. I mean if you are going to pay for a dedi for this not very likely to happen scenario, you can probably save two-thirds using a few VPSes and if the request comes in 6 months later, your savings will more than pay for a VPS to do the job, especially if you consistently work with a friendly provider who can probably do something to fit your needs.

    I actually have even paid the first month of my dedi, got a lucky transfer from some nice guy on the old forum. Yes, after coming to LE* I kinda became addicted in owning VPSes, I don't know why. Now that you've mentioned, the UltraVPS 4GB (now I'm sad I didn't get the 8GB one), Avoro 6GB and Contabo (I know some people hate it) 8GB might be a good setup. The only problem with the Avoro is that beside it's perfomance being good (at least in the paper), it lacks the space that UltraVPS (110GB) and Contabo (200GB) has.

    I'm gonna be 100% honest, the main reason I went for a dedi in the first place was to use Proxmox and have an easier time backing up stuff. I'm an average linux user, so manually backing up services on separate VPSes sounds like a pain to me. The other thing was of course the disk. Currently I only have one big disk VPS and I couldn't use as storage or I'd need offline backups.

    Considering those three VPSes I listed + 2 HH storage, I'd be saving only $54/yearly from the Dedi. If I'd keep the dedi, I would most likely have to keep at least HH 3TB and UltraVPS 4GB (for the really important services, like Bitwarden). That would increase the overall difference to something near $180/yearly.

    Just taking notes that the dedi price is currently 22EUR/mo. I recently abandoned another dedi, a $14/mo from drserver with 16GB, 1x2TB, 100mbps and the Atom C2750. I actually don't know, I guess the sb servers from Hetzner are really attractive.

    Oh, about the location, I'm actually in South America.

  • edited December 2019

    I would Keep the 22 eu hetzner and 1x3TB backup kvm.

    You can do a lot with just this setup.

    Edit: just saw your location. Keeping the drserver Dallas avoton dedi is suddenly compelling for vma backup storage and light self-hosting.
    Edit2: read again. You dropped it.

    Maybe just a 6.95 8GB smarthost nvme KVM in Dallas for anything high iops.(+ drop the EU vps? )

    Thanked by (1)Ouji
  • vimalware said: I would Keep the 22 eu hetzner and 1 3TB backup kvm.

    You can do a lot with just this setup.

    Yeah, I thought about that too. Only issue is that I have ALL MY PWD in my Bitwarden self hosted instance and I'm gonna be honest, I don't do export backups very often, so I rely a lot on it. Although I've noticed that I can keep using it for awhile even if the server is offline, I just don't know how long. If I were to have my dedi offline for a while, I'd need to spin my latest backup from the bitwarden instance into somewhere else (in case my offline bitwarden stopped working).

    I was gonna self host non critical mail as well, but after MXRoute deals I don't think I will go through the hassle.

  • Don't selfhost mail. Use the time saved to learn other skills

    Thanked by (1)chimichurri
  • @vimalware said:
    Don't selfhost mail. Use the time saved to learn other skills

    Yes, exactly. I gave up instantly after I started reading about it in depth. Not worth the time.

  • Also, I don't like the idea of online password managers.
    I use keepassxc (updated via snap) and sync it with my Seafile.
    I always have offline access to the last updated version.

    Thanked by (1)poisson
  • @vimalware said:
    Also, I don't like the idea of online password managers.
    I use keepassxc (updated via snap) and sync it with my Seafile.
    I always have offline access to the last updated version.

    I don't mind that, to be honest. I don't like the idea of needing to type multiple 20 character passwords if I happen to be on a place where I cannot access my Seafile instance. I feel like I need a certain balance between security and convenience.

  • There is an old proverb going long like this, never put all your eggs in one basket.

    And there is also another old proverb here, no use crying when all the eggs spilled.

  • @Ouji said:
    I actually have even paid the first month of my dedi, got a lucky transfer from some nice guy on the old forum. Yes, after coming to LE* I kinda became addicted in owning VPSes, I don't know why. Now that you've mentioned, the UltraVPS 4GB (now I'm sad I didn't get the 8GB one), Avoro 6GB and Contabo (I know some people hate it) 8GB might be a good setup. The only problem with the Avoro is that beside it's perfomance being good (at least in the paper), it lacks the space that UltraVPS (110GB) and Contabo (200GB) has.

    I'm gonna be 100% honest, the main reason I went for a dedi in the first place was to use Proxmox and have an easier time backing up stuff. I'm an average linux user, so manually backing up services on separate VPSes sounds like a pain to me. The other thing was of course the disk. Currently I only have one big disk VPS and I couldn't use as storage or I'd need offline backups.

    Considering those three VPSes I listed + 2 HH storage, I'd be saving only $54/yearly from the Dedi. If I'd keep the dedi, I would most likely have to keep at least HH 3TB and UltraVPS 4GB (for the really important services, like Bitwarden). That would increase the overall difference to something near $180/yearly.

    Just taking notes that the dedi price is currently 22EUR/mo. I recently abandoned another dedi, a $14/mo from drserver with 16GB, 1x2TB, 100mbps and the Atom C2750. I actually don't know, I guess the sb servers from Hetzner are really attractive.

    Oh, about the location, I'm actually in South America.

    Sorry about think you are in China. The currency is spiraling downwards too. :)

    Ok, now that I know your location, to keep everything that you have going at a really decent cost, I am going to suggest that you go with @HostDoc in either Kansas or Dallas. I believe that HostDoc supports nested virtualization for Proxmox (@dahartigan can confirm this). This takes care of that issue.

    Kansas City or Dallas will provide you much better latency and speeds to South America. Also HostDoc is a reliable provider, so that's definitely a super plus.

    Between the two, Kansas City is much more value for money because it uses older hardware, but if you are mostly doing self-hosted applications, the host node specs are good enough for you. Let's look at the pricing of their highest specced VPS in Kansas City on their flash sale page for £50 a year:

    • 4vCores @ 2.4 GHz
    • Unmetered @ 1Gbps
    • 8GB RAM
    • 1 IPv4 + /64v6
    • 50GB SSD
    • 50GB HDD

    This is definitely more than sufficient for your needs. If you don't want the hassle of having 50GB block storage, I am sure you can discuss with HostDoc to increase the SSD upwards to 100GB instead splitting the storage for a fee. All considered, you will likely pay less than half of the 22 euros a month going this route and meeting your needs too. The savings can more than pay for your HH storage, and even after this, you will still have some money left.

    I am with @vimalware on not using Bitwarden and using KeePassXC. This is my setup as well. You just sync your encrypted kdbx file using whatever way you like (Google Drive, Dropbox, Nextcloud, Syncthing, email yourself whatever you prefer) and you have access to your passwords with or without internet connection. There are lots of apps (mobile, PC and online) that are compatible so you don't need an always-on VPS to access your passwords.

    If you really want Bitwarden, I would say a cheap NAT OpenVZ VPS will do the job in terms of providing redundancy.

    Thanked by (2)vimalware HostDoc

    Deals and Reviews: LowEndBoxes Review | Avoid dodgy providers with The LEBRE Whitelist | Free hosting (with conditions): Evolution-Host, NanoKVM, FreeMach, ServedEZ | Get expert copyediting and copywriting help at The Write Flow

  • Haha. This could be a separate category : Tidying up: server edition
    Inspired by

    Thanked by (2)poisson Ouji
  • If a VPS does not spark joy, cancel it!

    Thanked by (1)chimichurri

    Deals and Reviews: LowEndBoxes Review | Avoid dodgy providers with The LEBRE Whitelist | Free hosting (with conditions): Evolution-Host, NanoKVM, FreeMach, ServedEZ | Get expert copyediting and copywriting help at The Write Flow

  • @cybertech said:

    @Neoon said:
    I would not suggest to put everything on one big machine.
    Because if this thing goes down, you gonna have a bad time.

    Better split stuff up.
    Even hole datacenters can loose power and/or network, we have seen such stuff multiple times.

    Even if you have 3 different power sources, you can still fuck it up, like OVH did.
    The hole DC went blyat.

    We need to have redundant datacenters then. Someone's gotta build one beside OVH Leaseweb Hetzner

    BAAAA, OVH had 3 redudant, independed power sources.
    Still it fucked, the MAGIC says: The more complex it gets, the harder it gets to fight against Morphies Law.

    Thanked by (1)uptime
  • Buy another dedi to add proper redundancy to your data, eventually picking different providers, datacenters and possibly countries;

    dismiss anything you won't possibly ever use unless you believe you need a point of presence in that country/region. You believe that, at some point, in the future, you'll need that stuff? Then you'll buy it when such a need arises. No need to idle it for months or years

    /me proceeds to idle hordes of VMs and dedis, "just in case"

    Thanked by (3)uptime Ouji K4Y5
  • Divide work into different levels, are they important?

    Action and Reaction in history

  • mfs said: dismiss anything you won't possibly ever use unless you believe you need a point of presence in that country/region. You believe that, at some point, in the future, you'll need that stuff? Then you'll buy it when such a need arises. No need to idle it for months or years

    I think that is the most important point that I'd need to take from your comment is if I'm looking to save money, I'd be saving more if I'm not paying for an idle service for months or years, like you mentioned. If I feel that I will ever need more stuff I'd just buy when needed, because no deal is gonna save more money or enough money to pay for the several months or years it will be idling.

    But looking into it, I might not need 32GB RAM now, maybe 16GB, but in that case, I don't think I will ever be able to get another 22EUR dedi with those specs, maybe another luck transfer (I requested this one in the old forum and got from someone not using). And I need to be honest that having more resources available will push me to host more stuff as I don't like the ideia of needing to buy more resources to host stuff, mainly because I'm not sure if I'm gonna keep what I initially because it might not be useful or I might not like the tool.

    That's a good takeway. A few of my invoices were cancelled again, but I think it's better to not have buyers remorse.

    elliotc said: Divide work into different levels, are they important?

    Most things are not that important, but I might get new stuff that are important.

  • Ouji said: 22EUR dedi with those specs, maybe another luck transfer (I requested this one in the old forum and got from someone not using).


    On HB I shortly considered to transfer away exactly one of those boxes before BF, you've missed that I guess... I did announce it only a few days before BF (I had high hopes to... consolidate a couple of boxes into a shiny AX server if only Hetzner chose to waive fees or something)
    It's still in my hands, I may page you if I'm considering to let it go in the next month or two if you want

    Ouji said: A few of my invoices were cancelled again

    “Train yourself to let go of everything you fear to lose.” - Yoda

    Thanked by (1)Ouji
  • havochavoc OGContent Writer

    That's a lot of servers @Ouji

    I'm planning on consolidating with docker configs on larger VPS. Makes sense to have it consolidated to maximize each tasks ability to peak temporarily in that larger resource pool.

    Basically a 3vcore NVMe KVMs on each side of Atlantic. One to host public facing stuff, the other for special project purposes.

    All the TB heavy stuff I host on a home server, hence storage light NVMe.

    Thanked by (1)Ouji
  • don't keep all your eggs in the same basket

  • Have backups.
    In the low end segment, cheap vps/dedi + regular tested backups beats any redundant setup.

    Thanked by (1)Ouji
  • mfs said: On HB I shortly considered to transfer away exactly one of those boxes before BF, you've missed that I guess... I did announce it only a few days before BF (I had high hopes to... consolidate a couple of boxes into a shiny AX server if only Hetzner chose to waive fees or something)
    It's still in my hands, I may page you if I'm considering to let it go in the next month or two if you want

    Oh, that's nice. Mine is in Finland, if yours is in Germany it might have a better routing here (the routing is still average I'm pretty far from EU). I actually actively requested a server and got a reply in less than 1 day. The guy even gave the server like 8-10 days before the next invoice, so I got a few days to test everything out.

    @havoc said:
    That's a lot of servers @Ouji

    I'm planning on consolidating with docker configs on larger VPS. Makes sense to have it consolidated to maximize each tasks ability to peak temporarily in that larger resource pool.

    Basically a 3vcore NVMe KVMs on each side of Atlantic. One to host public facing stuff, the other for special project purposes.

    All the TB heavy stuff I host on a home server, hence storage light NVMe.

    Yeah. I got a little impulsive thinks to LE* and now I'm trying to recover from the disaster that this year was (in terms of virtual servers and such).

    Thanked by (1)mfs
Sign In or Register to comment.