#docker

XedonTechnologie well if the image changed, they prob changed smth in how it starts - see logs, maybe they added a new option for it or so... ((some people like to specify excaly what IP to bind to - when say having a container that sits in 2 networks - but "service" should only be acessible to one of them ((as there is "no firewall inside docker networks"))

nwe well, the error you pasted clearly shows it attempting smth with "dial tcp 127.0.0.1:5432" - so maybe another var to set appart from CONCOURSE_POSTGRES_DATA_SOURCE?

and --link IS DED.

very starnage, the docker build has skipped the step of copying my config file

nwe ((thats also odd btw, one one hand "link" would make it the "localhost" yes - but you already give it the hostname in CONCOURSE_POSTGRES_DATA_SOURCE...)) -- so tldr, `docker network`create&attach (or --network in the `run` is "the real fix" -- and wrap that up into a docker-compose, noone wants 600char long shell commands :P

I'm banging my head over a good ol' standard_init_linux.go:190: exec user process caused "no such file or directory", most likely my dockerfile screwing up somewhere. Is there a way to know what's missing ?

lyr got a workdir set? - also dockerfile pls

tabakhase thanks for the help, though my question was more about is there something like --debug flag for a decent troubleshooting

lyr first line of /cozy/bin/entrypoint.sh is? (shell?) - i think alpine trying bash yields that exact error...

oh

hell

really nice catch

lot of alpine images just have that symlinked, some have actual bash, - but "latest" may be pretty vigin...

speaking of - never use latest... its "junk" unless you know its not - it may be random ci artifact that failed every test there is and such ;D

I know the pro & con

thakns for help tabakhase , turns out someone messed with the Dockerfile and commited the last version with missing a "COPY" line to copy the config.

XedonTechnologie neat :P

Hi there, complete n00b question: I have a 6 node swarm all nodes can be seen with $ docker node. Is running 18.06.1-ce & 17.05.0-ce an issue? I can't seem to deploy Portainer agent on nodes -05 & -06 running 17.05.0-ce engine

*$ docker node ls

How does a docker-compose memcache not have its hosts written into containers on the same network?

Are there any notable performance difference between a bind mount vs a volume mount?

"depends" ;-)

hehe :D under what condition does it depends?

i think bind is technically "faster" - as it can skipp all the layering & copyOnWrite gezizzle

I like volumes, since they are named and have a central registry. But they are less accessible than bind mounts and is more tricky when it comes to top-level permission

not 100% if thats only for writes, reads should be fairly transparent on both ends i think

so volumes do have cow-mechanisms set in place?

how do I check the logs for a crashed container? I started it with docker-compose up

well thats what all those layering filesystems do or not? o.O :D

docker-compose logs

and command in docker-compose overrides the one in the docker-file right?

its entrypoint which is harder to override

drive seems mostly to "not docker" for super-io (database server is one thing you often find on "bare" still for example) - and for actual "docker stuff" volume or bindmount is kinda ussage question -- "to just deploy anywhere anything a fixed build" you would not want bind, even if a bit faster - rather scale horizontally :P

especially when you then dunk into managementsystem,s alla kube and whatanot where volumes are not /var/lib/docker :P

ixxie: yes, it overrides. Why would entrypoint be harder to override?

ixxie you can overwrite both in all places - but they "are not rly the same" - they overlapp for a few things sure and one can use entrypoint as cmd, but generally they are more like "delayed build" ((see canidates ending with e`xec "[email protected]"`)

but thats still just 2 of like 4 patterns how one can build dockers

take php-docker as random example -- `docker run php:cli -r 'echo "mo";'` works, as does `docker run php:cli php -r 'echo "mo";'` --- its entrypoint "checks if it starts with a "-" and injects php on its own

tabakhase: if a command is specified in the dockerfile and then another in the docker-compose, will the first one be executed at all?

nope

that is what I am hoping for

its more acting as "default value" than a "stacke d thing"

wanting to run multiple commands sounds more like a resign error ;-) tell me more... --- or you want a entrypoint ;D

s/resign/design

I want to override

weirdness because I try and run `command: /bin/sh` and the container crashes

(("the book" says ONE process per container - and that "likely" should be whats in CMD in the end... -- and while there are valid exceptions to that rule, they should be done "right" with say a taksmanager like runit, s6, dumyinit and so on ((cause "just kicking it into the background" will never ever deliver your "kill" command from the outside for example)

or maybe not crashes because exit code is zero

hers a bit more reading on "the exceptions" http://phusion.github.io/baseimage-docker/ -- dev-env with "all in one dockers" or the first dockering projects is where stuff like that comes super handy

I just want to run the same container without starting the service so I can run tests on it

sh ends with "nothig" if you give it nothing todo

whats the best way to keep the container up without doing anything special?

why "without service"? --- and running tests is usually more an "exec" thing

I thought /bin/sh would start a shell