|
i'm trying to improve my leet container skills by learning about etcd but nobody seems to know just what it might be used for all the tutorials i can find say it maintains a distributed key/value store but how does that come up in a practical system? how do values get initialized/changed from outside the cluster? i dond't give a gently caress what consensus algorithm it uses, why even mention that? help me yospos, what have you used etcd for irl
|
# ? Dec 27, 2016 18:38 |
|
|
# ? May 6, 2024 14:01 |
|
comedy answer: it's good for writing obscure tutorials and blog posts
|
# ? Dec 27, 2016 18:38 |
|
suck it off
|
# ? Dec 27, 2016 18:39 |
|
put your dick in it
|
# ? Dec 27, 2016 18:39 |
|
idk op, it's a logging thing I guess
|
# ? Dec 27, 2016 18:44 |
|
I think the keys/values are supposed to be used to manage configs inside the container (e.g., inside the /etc folder). but I've never used it because I've never used a container in production.
|
# ? Dec 27, 2016 18:47 |
|
service discovery and config mostly. probably anything else you might use zookeeper for.
|
# ? Dec 27, 2016 18:50 |
|
JimboMaloi posted:service discovery and config mostly. probably anything else you might use zookeeper for.
|
# ? Dec 27, 2016 18:58 |
|
gee whiz this is as bad as facebook's anti-sales-pitch. "it keeps you connected" yeah yeah but what can I DO with it
|
# ? Dec 27, 2016 19:01 |
|
yeah like 30 seconds of looking at it makes it sound like yet another "distributed KV, but for config poo poo so its super lightweight whatever the gently caress that's really supposed to mean!" basically in some cluster environments you might want a service that provides configuration and service endpoint information for members of the cluster to look up so they can do whatever job they are supposed to do (like Apache HBase likes to keep information about where Region Servers are and stuff in Zookeeper) of course to avoid having a single point of failure, this service itself must now be a distributed system, but is also very important for it to be fault tolerant so you spend all the time and resources on that can of worms as well, etc etc
|
# ? Dec 27, 2016 19:24 |
|
so, for an example a yosbbs prototype was on a docker/coreos cluster i had four little bbses running side by side. load balancers aimed inbound connections at them. All of them needed to hit a specific memcache docker, which could move around so i set up the memcached to announce its ip in an etcd store, and then the bbs was modded to just pull that value from that key whenever it needed the memcache it was pretty cool tbh
|
# ? Dec 27, 2016 19:27 |
|
basically if you want some dynamic information that can be read/written by members of your cluster, stuff like zookeeper/etcd are useful, but its not especially necessary for static config stuff
|
# ? Dec 27, 2016 19:34 |
|
did you suck it off?
|
# ? Dec 27, 2016 19:34 |
|
btu what if the internet goes out?
|
# ? Dec 27, 2016 19:47 |
|
I'm in the health industry, and we use windows and citrix with some legacy ibm mainframe stuff. what would etcd bring to this?
|
# ? Dec 27, 2016 19:48 |
|
Annual Gift Panda posted:I'm in the health industry, and we use windows and citrix with some legacy ibm mainframe stuff. what would etcd bring to this? etsy DEEZ NUTS
|
# ? Dec 27, 2016 19:53 |
|
Annual Gift Panda posted:I'm in the health industry, and we use windows and citrix with some legacy ibm mainframe stuff. what would etcd bring to this? more things to janitor basically whenever you bring clusters into things it means more janitoring
|
# ? Dec 27, 2016 19:56 |
|
how does this differ from redis
|
# ? Dec 27, 2016 19:59 |
|
Jonny 290 posted:so, for an example so its like a second memcached that tells the servers where the first memcached is
|
# ? Dec 27, 2016 19:59 |
|
i have the suspicion that computer men just invent these new systems and keep stacking them up just out of boredom
|
# ? Dec 27, 2016 20:00 |
|
Bloody posted:so its like a second memcached that tells the servers where the first memcached is sort of, but the replication/consistency's different
|
# ? Dec 27, 2016 20:07 |
|
Bloody posted:how does this differ from redis as far as I remember redis fault tolerance isn't the greatest you also wouldn't necessarily store the kind of information in one of these systems that you would store in redis (and they don't really offer the data typing that redis does), in fact a great way to break zookeeper is to try to store something over this data limit (like 1MB) there's a certain amount of physical architecture to consider with systems like zookeeper/etcd as well Bloody posted:i have the suspicion that computer men just invent these new systems and keep stacking them up just out of boredom yeah this in ways
|
# ? Dec 27, 2016 20:09 |
|
Jonny 290 posted:so, for an example
|
# ? Dec 27, 2016 20:19 |
|
If Consul fixed the issues with its Raft protocol stuff just sometimes totally making GBS threads itself it would be pretty dec for managing situations where you got a lot of instances of various services spinning up and down all the time. Like if you were actually using AWS at scale the way it's envisioned to be used. However, if you're like 99 percent of people that never operate near the scale where that architecture is necessary then just use Puppet and make your life way easier.... unless you started using Puppet before 3.x in which case you're doomed sorry
|
# ? Dec 27, 2016 20:22 |
|
oh no blimp issue posted:put your dick in it sounds like a real clusterfuck haha lol
|
# ? Dec 27, 2016 20:23 |
|
Gazpacho posted:sounds like a real clusterfuck haha lol
|
# ? Dec 27, 2016 20:27 |
|
it's alright for locking and leadership. i use it so that only a single emr job can write to a data warehouse at once because multiple gigabyte data loads happening concurrently tank performance
|
# ? Dec 27, 2016 21:02 |
|
Gazpacho posted:sounds like a real clusterfuck haha lol
|
# ? Dec 27, 2016 21:24 |
|
https://curator.apache.org/curator-recipes/ pretend you want to do anything on this side except you hate java so much you'd rather use a less-featured half-baked buggy randomly deadlocking babbys first kv store because it's written in go (tbh its probably fine now but boy did goraft suck when i looked at it 2 years ago)
|
# ? Dec 27, 2016 21:58 |
|
Bloody posted:how does this differ from redis redis clustering is loving garbage
|
# ? Dec 27, 2016 22:48 |
|
unless you want to buy redis labs thing
|
# ? Dec 27, 2016 22:48 |
|
Gazpacho posted:what might i use zookeeper for jfc lol
|
# ? Dec 27, 2016 22:50 |
|
we don't have anything deployed in containers yet, it's in the cards though the systems we're looking to containerize are all configured with manually edited yammel files because
|
# ? Dec 27, 2016 22:55 |
|
oh no blimp issue posted:put your dick in it Extremely this.
|
# ? Dec 27, 2016 23:02 |
|
i wrote up a whole plan recently to use AWS devops services to manage updates of our service clusters so that we don't have to bake images by hand and sent it off to the relevant party. the answer came back "beep boop vendor lock in is inevitable, docker is always the answer"
|
# ? Dec 27, 2016 23:28 |
|
that's fine, AWS hosts docker all day long. (I know, i know.)
|
# ? Dec 27, 2016 23:30 |
|
has anyone said ABSOLUTELY NOTHING yet
|
# ? Dec 27, 2016 23:57 |
|
it provides key-value pairs for configuration parameters, like you might store in /etc, but as a daemon. what's to understand
|
# ? Dec 28, 2016 06:09 |
|
JewKiller 3000 posted:it provides key-value pairs for configuration parameters, like you might store in /etc, but as a daemon. what's to understand
|
# ? Dec 28, 2016 06:16 |
|
|
# ? May 6, 2024 14:01 |
|
if you don't understand a basic description like "etcd is a distributed key value store that provides a reliable way to store data across a cluster of machines" then i don't know how anyone in this forum or on this earth is going to help you learn your job
|
# ? Dec 28, 2016 06:18 |