Question: anyone please tell me how to do two container HA in the same VM?

Asked By
tovadavid
Asked At
2018-04-22 09:06:01

Found 15 possible answers.

User Answered At Possible Answer
coderanger 2018-04-22 09:46:21 Not sure what you mean, HA usually means physical isolation since if one goes down, the other should not
tovadavid 2018-04-22 09:50:01 can i do HA for 2 containers? if one down, the other is still up.
coderanger 2018-04-22 10:02:16 What does that even mean? Most HA plans use multiple hosts with increasing levels of separation in the hopes that one failure won't affect every HA plan. As such, the actual availability benefit is near zero (kernel panic, hardware failure, networking problem, I/O error, etc) Technically yes, but almost anything that takes out one copy will also take out the other That's like asking if you can run two copies Postgres on the same server "just in case" Containers are just a fancy kind of process
ouhab.idir 2018-04-22 12:13:58 I’d like to share this with you: https://medium.com/@ouhab.idir/devops-or-not-devops-that-is-the-question-8e59dbfb3944
acad.sergio 2018-04-22 16:30:15 Thanks, @rlister your simple descriptions, and wisely advice seems to come from use in fact. I’m a noob too, working simple EC2 instances and simple CI system ( https://buddy.works/ ) that runs a series of SH’s to “deploy” from Github branch hooks. I will keep dirty my hands now, but what can be the next step to evolve this process?
geekman2 2018-04-22 21:00:41 How common/expensive is a Java+Ops skillset ?
ulan 2018-04-23 03:32:44 if you're company has multiple products would you deploy them all in one k8s or swarm cluster?
coderanger 2018-04-23 03:45:51 @ulan Generally rule for k8s is one cluster per failure domain, and proooobably put dev and staging on a separate cluster from prod
ulan 2018-04-23 03:46:48 what do you mean by failure domain? so different cluster for domain1.com and domain2.com ?
coderanger 2018-04-23 03:48:28 failure domains are the general form of what what Amazon calls "availability zones"
jhuizy 2018-04-23 03:48:32 i’ve thought about playing with the idea of multiple services in a single cluster to reduce costs for low traffic/risk projects such as POCs/MVPs and personal projects
coderanger 2018-04-23 03:48:57 i.e. spaces that are distinct and isolated so a failure of one is less likely to result in a failure of another nearby failure domain But in general yes, you would run many services in a single k8s cluster, that's kind of the point Each cloud calls them different things and has slightly different availability promises/isolations (and if you are building your own hardware, it's up to you to decide exactly what it means) Hence, failure is isolated For example, each AZ in Amazon runs on their own power circuits so if something goes wrong with the power systems, it will only affect that specific AZ
ulan 2018-04-23 03:51:23 I see... I kind of get it for dev but what's the point of separating out staging in a different cluster though, ? if it should theoretically be identical to prod.
jhuizy 2018-04-23 03:54:40 staging should be an entirely separate “stack” built with the same config as prod
coderanger 2018-04-23 03:54:57 Because k8s tools (and humans) are very new in most ways and the chances of an "oops" scenario are high and usually I've seen people throw dev on the staging cluster because it's easier And yeah, prod and staging are usually similar (except in scale)

Related Questions