Failsafe service?

Question

PHP

Hello!
There is a server with 100+ clients. Clients sit on a common code base, while each client has its own database (MySQL). From time to time, force majeure occurs at the hoster, DNS service, etc., while there is an interruption in the provision of the service, which negatively affects our
reputation one, we could easily switch to duplicated services (ideally, this would happen automatically.) If at the same time it would be possible to do
load balancing, then it would be even better. not
a panacea and there may be desynchronization and other problems.
Tell me, is it possible to come up with something similar on cloud services? For example,
Amazon or maybe someone else?
How will there be duplication of services? How should the architecture be built?
How reliable is it, what are the pitfalls?
Thanks for the help!

Answer 1

Are you by any chance tw1.ru? And then, with sad constancy, errors fall out ...

Answer 2

about Amazon
habrahabr.ru/post/147390
habrahabr.ru/post/148240

Answer 3

IMHO look for a specialist who will help you, if you want yourself, then :
en.wikipedia.org/wiki/%D2%E5%EE%F0%E5%EC%E0_CAP
habrahabr.ru/post/136398
systems in the group of companies)

Answer 4

Well, MySQL more or less normally knows how to multi-commit-master (in the form of Galera). It will only be slower and you need at least 3 nodes in different places, otherwise it will go away with split-brain.
As for the rest, [email protected] is right - a complex topic is complex, specialists who know how to do it receive a decent amount of money and have long been employed for 20 years in advance.
I can also give consultations in gabber-mail for free, so that you have a different point of view.

Answer 5

replication solves
no panacea

Answer 6

Tell me, do I understand correctly that we just take vps on Amazon? I assumed that Amazon provides the infrastructure and takes care of the availability of vps in the event of a crash on the server on which the vps is located.

Answer 7

Then two questions - did you have experience in building fault-tolerant systems on Amazon and the same solutions, but using your own servers? What are the pros and cons of the options?
How much will such services cost?

Answer 8

Use regular Master-Slave replication to back up the database (only the Master will work with applications). In case of failure of the slave - it is rebuilt in the background. If the Master exits, you switch to the Slave. It is better to use 1 Master + 2 Slaves.