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
@miu
Without knowing anything but a rather superficial description ... and with the (purely subjective) caveat that I generally don't trust MySql and always prefer Sqlite (for small stuff, which is adequate for much more than most think) or PostgreSQL for big stuff ...
I do not think that a dedi is needed. A good (and powerful enough) VPS not only can also run a DB server but in fact it will often do so better because the hardware and configuration behind a good VPS usually is actually better than an average dedi. So unless your DB is driving a really large and busy site running it on a good KVM VPS will often be actually better and much, much cheaper.
Re the location one can easily err on the wrong side. Keep in mind that on the internet it's not about distance (in miles, km) but about latency. So obviously 2 (virtual or dedi) servers on a MAN would be a good solution but I'd rather suggest a slightly different solution as availability seems to be of concern to you: I'd look for 2 VPS from different (but good) providers in two cities (not too far apart) like, say FRA and AMS or AMS and LON who have really good connectivity/low latency and who (or at least whose data center) have a direct FRA/AMS or AMS/LON IX connection.
My personal choice (which might not be the right one for you) would be to get an adequate (4 vCores, 4 GB, SSD RAID or whatever your situation asks for) KVM VPS from @Clouvider (LON, in the IX) and some good german or NL provider in a DC with IX (Amsix or DeCix) presence. Looking at DeCix or Amsix it should be easy to spot a provider with a sound presence who has a good reputation. One reason for that choice is that in addition to 2 providers in 2 data centers in 2 different regions I'd also have those servers in 2 (at least somewhat different) jurisdictions.
For something I deem important I would also have a storage VPS with one of the providers or, if available, a solid backup solution.
Please kindly note that while I'm usually ready to help out with advice I also like to stay on topic in a thread, so I suggest you contact me by PM for anything not directly related to "Realtime File SYNCHRO betw 2 servers (separated datacenters) in master/master mode".
P.S. Except for very few extreme cases I would not bet on SQL server replication but handle that mayself in my application code. After all I know best what data are written and read how frequently, how stale or fresh they can/must be, etc, plus I like to tune caching just the way I deem right for a given use case. And it's usually easy too; just a couple of lines of code plus I can play with compression, crypto etc. just like I feel to need it.
Regarding MySQL, look into Mariadb Galera Clusters.
I actually disagree with this. SQL replication is a thing that is used pretty often - and as such, great software options exist for it (like aforementioned Galera Cluster). I would never attempt to DIY SQL replication myself.
DRBD?
+1 for MariaDB.
i also do not remember issue with MySQL latest 10 years (from time when i'm using it), but on other side i must admit do not have experiences with POstgre yet ever, so i cannot objective & relevant compare them and say what is sure better and more reliable (for regular use at middle or high difficult dynamic websites/webapps or their DB replication)
SQLite i'm using only for a small websites, also do not remember problems with lite..
and for completion and better imagines (what i really had on mind when started this thread) this was my original idea, wishes before "ground me" from notes by FHR and jsg, what i sketched on following scheme picture:
one more time many thanks for all relevant contributors here for their all already posted replies
and a few clients from them are operating 24/7, and direct depending on their online services, so required is best possible uptime / high availability, not very tolerated and problem is already time more as 5 minutes of outages and a few times yearly (2-3x) only (this is period basically for reboot less-more).
Many times, people wants to have rocksolid stable services w 100.00% uptime and online so long without outages as year is long, but are most time not willing pay for it extra because they does not know how difficult such thing is in realilty, Due from this reason solution modify and tune their websites for properly and exact work with multiple mirror nodes as was suggested here is not very applicable solution (do it for each one extra and unpaid, and spent a lot time and work free on it), very nice would be to have "uni HA autofailover platform" when u can then simple move all critical clients (who are sensitive for each outages / working 24/7), as one solution w least time and work consumed able resolve them all at once, without resolving each one separately and extra. But unfortunately after discussion in this thread i now know this it not easy and questionable how reliable, well and with how level performance can such anything work ever... but at the end i can try it on the fly for test
if i good remember i remember a few months back was whole FR OVH datacenter (or part) offline more than hour (or longer)... and also best of the best providers haves outages time at time (caused by local network/datacenter, not caused by node/server level issues).
So i think - to have only very reliable HA cloud instances (mean real cloud instances w not internal HDDs but with FibreChannel connected redundant SAN storage, openstack etc. cloud solution - what are able move whole VPS instances & IPv4 without reboots on other physical node if necessary, real cloud providers such as DOC, Vultr, Prometeus, OVH etc..) at all still can be not enough to achieve and cover maximum possible uptime, and if u have "luck" still your VPS can be working but several tens of minutes or hours offline from internet (caused by datacenter unexpected issues, DDos attack on single point/your only IPv4 etc. if you are using only one location and datacenter)
BTW: any recommendations or suggestions for stable providers who offer with VPSes also internal network between their location (min. 2 loc), ideally in EU:
i can start with what i know:
(DOC, Vultr and Linode offers private LAN unfortunately only within current/one datacenter, not cross datacenters.. Hetzner also only within DC if i'm right)
I would just do OVH GRA and RBX tbh.
Hotel echo tango zulu november echo romeo.
I'd just set up a VPN tunnel between the locations (eg. using Tinc or ZeroTier).
i contacted linode support w ticket related to this matter and also did ask for approx latency (FRA - LON - US) and they also mentioned VPN for connection between locations, and replied me following:
So i think couple from DC LON-AMS or AMS-FRANK or LON-FRANK by Linode, DOC or Vultr can be maybe also usable (except OVH GRA RBX) for MySQL cluster (two replicated MySQL servers in master/master mode) to a certain level of load MySQL, usage (says low and middle).
But probably / seems to be OVH as best, fastest and uncomparable most reliable option (thanks to their possible private network connection through their own internal network directly)
No problem. You are right for many many cases. The reasons I don't like it (except for aunt Lizzies hobby cake shop) is (a) generalized vs specialized solution which always brings along some trade offs, and (b) it can get ugly quickly with write through mem tables, priority and speed hierarchy, and generally more complex configurations. Plus, but that's my personal issue, I do not trust their safety and security.
But again, for the majority of linux/MySql/PHP/ngingx|apache installations out there it'll do just OK.
What's wrong with Dropbox ? I believe it has Linux native client