just fyi: thats the discussion on `#couchdb` on freenode describing what happend to me
``` <ufobat> i forgot to set the -setcook in my vm.args file, when i set up a cluster. funny thing that happend, my cluster worked first. but afer a node went down and came back up, my cluster didnt work anymore <ufobat> my question is, why did it work after the initial setup of the cluster? <rnewson> in the absence of that setting, each node will have generated and stored a random value to ~/.erlang.cookie <rnewson> they obviously won't match across the cluster, which is why the cluster doesn't work <ufobat> so it shoud not have worked ever? <rnewson> well, if you used the cluster setup tool we ship, it may have set the cookie for you <rnewson> anyway, it's easy to solve. add the -setcookie line to all nodes with the same value <ufobat> i used fauxton <rnewson> it's nowhere near as annoying to fix as getting -name wrong <ufobat> but there is no cookie set in the vm.args <rnewson> then each node is reading from a file <rnewson> erlang 21 (I think) finally removed that fallback behaviour <ufobat> then another question is raising in my mind: when you need to set both the nodename and the cookie, shoudn't it be possible to be set both in the apache/couchdb-docker image <ufobat> currently you can only set the nodename: https://github.com/apache/couchdb-docker#detailed-configuration <rnewson> I know approximately nothing about docker, sorry. <rnewson> besides an instinct not to use it in production, which probably dates me. <ufobat> well. you can pass the nodename at (the first) startup time, you can not set the erlang cookie. imho that sounds at least strange to me <rnewson> possibly a reflection of the fact that you might care what the names are, but you don't care what the cookie is so long as the nodes agree and that's it a big random value <ufobat> i see. that i missed the point how the instances of the cluster agree on that.. no actually i missed where those instances store it. since it was obiouslie not vm.args <rnewson> ~/.erlang.cookie I think, that's just normal erlang behaviour. <ufobat> ya sorry you wrote it. <ufobat> nope there is no .erlang.cookie on my couchdb clouster instance docker container :( <rnewson> hm <rnewson> is it just a single node cluster? <ufobat> no it consists of 2 nodes <ufobat> and i am absolutly sure it had worked in the beginning because i've created a db on node1 and asked for the db on node2 <ufobat> which was not longer possible after rebooting one node <rnewson> hm <rnewson> nowhere in https://github.com/apache/couchdb-docker do I see the cookie being set, the word cookie isn't even used <ufobat> i am just wondering.. i've got a erlang cookie now, and rebooting a node leads to the expecte behaviour <rnewson> I would add the -setcookie line to vm.args of both nodes and restart them <rnewson> and maybe file a bug against that repo for not writing the cookie into the vm.args for you <ufobat> i just created a issue and asked if that shoudn't be set somewhere: https://github.com/apache/couchdb-docker/issues/106 because i know that there should no more environtment settings be introduced <ufobat> ty for your help again, rnewson :) <rnewson> np ``` [ Full content available at: https://github.com/apache/couchdb-docker/issues/106 ] This message was relayed via gitbox.apache.org for [email protected]
