<roidrage>
it's just http as rubygems barfs on the missing certificate
<roidrage>
but it's a start. it basically redirects everything directly to s3
pipework has joined #rubygems
dvu has quit [Ping timeout: 256 seconds]
nicek has joined #rubygems
<nicek>
hi
<samkottler>
roidrage: what's your plan for telling users to use your proxy?
<roidrage>
i don't have one yet :)
<samkottler>
nicek: hey, what's up?
<roidrage>
i suspect with the source now being a direct url there's no easy way to sneak it in
nicek has quit [Remote host closed the connection]
<roidrage>
the general downside is that the source is tied into the Gemfile.lock
<roidrage>
need to figure that out
<roidrage>
i also started adding support for retries to bundler to mitigate the issue on that end as well
ZachBeta has quit [Quit: Computer has gone to sleep.]
fromonesrc has joined #rubygems
kyzh has joined #rubygems
kyzh is now known as Guest98420
karmi has quit [Quit: Leaving.]
huoxito has quit [Quit: Leaving]
karmi has joined #rubygems
<roidrage>
samkottler: but if you have any suggestions, let me know :)
<samkottler>
roidrage: I like the idea of retries, especially if they are local
<samkottler>
so for example, if a gem install times out, it'll get re-executed n number of times
kgrz has quit [Ping timeout: 245 seconds]
<roidrage>
yeah
pipework has quit [Remote host closed the connection]
<samkottler>
where n is some number low enough that bundler doesn't DDoS rubygems.org :)
<samkottler>
3 probably feels like a good default
mootpointer has joined #rubygems
pipework has joined #rubygems
<karmi>
samkottler: Any idea about adding "vagrant" to the Gemfile for rubygems-aws? Getting `already activated` errors without it...
<samkottler>
karmi: that will break with newer versions of Vagrant because it's no longer distributed as a gem
Elhu has joined #rubygems
<karmi>
samkottler: I know, but for older ones, it might be a good thing.
<karmi>
samkottler: would it then "shadow" new vagrant versions installed from the package?
<samkottler>
karmi: that'll cause some issues :(
Guest98420 has quit [Ping timeout: 272 seconds]
<roidrage>
samkottler: yeah, i prepared some code for that, need to wrap it up in a nice pull request :)
<roidrage>
samkottler: do you have any thoughts/ideas on where the current ssl config for nginx falls short?
Noma4i has quit [Quit: Computer has gone to sleep.]
<dwradcliffe>
karmi: where are you getting an error?
<karmi>
dwradcliffe: hey! without `bundler exec`?
<dwradcliffe>
karmi: I'd suggest removing the gem and installing from package
<samkottler>
roidrage: I haven't hadn't had time to really dig into it yet, unfortunately
<roidrage>
okay, cool
<roidrage>
i'd love to help out with the ops part on rubygems.org at some point if there's interest
pipework has quit [Remote host closed the connection]
<samkottler>
roidrage: that'd be awesome! it's definitely a WIP right now
<dwradcliffe>
roidrage: ++
<karmi>
dwradcliffe: Sure. Can't do it now, though, I depend on the gem version in a cookbook I work on.
<karmi>
dwradcliffe: need to be able to require "stuff", no way to do it easily with the packaged Vagrant version
<karmi>
roidrage: hi! I'm working on adding Elasticsearch currently.
<karmi>
So, `dbmaster` fails now with:
<karmi>
* Error: /var/lib/pg_data is not accessible or does not exist
<roidrage>
karmi: hey! elasticsearch to rubygems.org for search?
<samkottler>
karmi: would you mind sending the email you sent earlier out to the rubygems.org list? there are some people on there who would probably like to help out
<dwradcliffe>
karmi: haven't had time to refactor yet :(
<karmi>
samkottler: I guess in this stage I need a realtime advice...
<karmi>
samkottler: I've spent a lot of time debugging it already, and these are just bits and pieces
<samkottler>
karmi: that address being hardcoded is technical debt
<samkottler>
it should just be in the node attribute
<dwradcliffe>
karmi: I have a branch that attempted to fix some of this
<samkottler>
a node attribute**
<karmi>
OK. I was wondering about some general issues in the whole setup… could be a bit improved here and there.
pipework has joined #rubygems
<samkottler>
karmi: yeah, we built it in 2 sleepless days so there is definitely some debt :)
<karmi>
I can imagine :)
<karmi>
Some time ago, me and my colleague prepared a demo full-stack Chef install, it uses Chef Server, but some ideas could be lifted from it, such as restoring from backup etc, https://github.com/karmi/chef-hello-cloud/
<karmi>
dwradcliffe: samkottler: what about the `/var/lib/pg_data is not accessible or does not exist` error?
<karmi>
That's the path set in the `rubygems_db_master` role. I guess it should be created by the pg recipe?
<samkottler>
karmi: yeah it should be when running on vagrant, in production that's an EBS volume
<karmi>
samkottler: OK, but why I get the error then...? :)
<karmi>
roidrage: You'd be interested in helping with the Elasticsearch integration?
<karmi>
roidrage: Lots of stuff to be decided IMHO, such as if to plan for backup/recovery, how to do it, etc etc.
<dwradcliffe>
karmi: can you post a gist of the error context?
<roidrage>
karmi: my personal experience with ES is rather limited :)
<dwradcliffe>
karmi: I've seen that before. If you ssh in and restart postgresql does it restart ok?
cowboyd has quit [Ping timeout: 256 seconds]
kgrz has joined #rubygems
wayneeseguin has quit [Ping timeout: 264 seconds]
sj26 has quit [Ping timeout: 245 seconds]
<karmi>
dwradcliffe: will try
d_roge has quit [Ping timeout: 256 seconds]
fromonesrc has quit [Ping timeout: 264 seconds]
sj26 has joined #rubygems
d_roge has joined #rubygems
wayneeseguin has joined #rubygems
fromonesrc has joined #rubygems
<karmi>
dwradcliffe: Sadly, no luck:
<karmi>
vagrant@rubygems-org-dbmaster:~$ sudo service postgresql restart
<karmi>
* Restarting PostgreSQL 9.2 database server
<karmi>
* Error: /var/lib/pg_data is not accessible or does not exist
<karmi>
Doesn't matter if I kill the process etc, it won't start, `cap chef:dmaster` fails on it as well.
<dwradcliffe>
karmi: does the dir exist?
<karmi>
dwradcliffe: No :) But I'm surprised by that — though I must be doing something wrong, this prevents the stack installation for everybody then.
<dwradcliffe>
karmi: well I'm not 100% sure the whole stack install works right now
cowboyd has quit [Read error: Connection reset by peer]
<karmi>
dwradcliffe: it's pinned in Cheffile, ref: "0.8.0", I think going with "master" would be safe enough?
<dwradcliffe>
yeah just yank the ref
* dwradcliffe
will brb
workmad3 has joined #rubygems
<karmi>
dwradcliffe: samkottler: Also, the vagrant vms use Chef 10.x, I think it would be wise to use a Chef-Solo Vagrant provider integration to upgrade it to newer version?
<karmi>
samkottler: About using node attribute in the dbmaster role — which IP would you use? node.ipaddress? node.network.xxx ? what about using private IPs in EC2?
<dwradcliffe>
I wouldn't just copy/paste, but you can do something similar to set the attributes in a recipe
<karmi>
dwradcliffe: yeah, that's a way to some solution
<karmi>
dwradcliffe: though the values set in the databag will override this
<karmi>
dwradcliffe: random other stuff I bumped into — the "user" and "sudo" cookbooks will wipe out any default user/sudoers/sshkey you have on the target machine. so eg. to test deployment against EC2 I had to run a databag for me. this feels wrong to me.
cowboyd has joined #rubygems
mockra has joined #rubygems
<karmi>
dwradcliffe: Also, why use Capistrano for the deploy, with all the overhead it comes with? Why not simply git fetch, reset, app restart?
mockra has quit [Remote host closed the connection]
<dwradcliffe>
karmi: we don't currently use capistrano for the app deploy
<dwradcliffe>
it's part of the chef process, although we have been talking about changing that
mootpointer has joined #rubygems
<karmi>
dwradcliffe: I abandoned cap based deploys long ago. Git etc makes it more lightweight.
<karmi>
dwradcliffe: Right now stuck on migration fail again — I think it's something about database.yml not being symlinked/copied
<karmi>
dwradcliffe: And this is too hard to debug with Cap.
<dwradcliffe>
karmi: I'm open to talking about changing that but probably not right away. There are bigger problems to deal with first.
cowboyd has quit [Ping timeout: 252 seconds]
<karmi>
dwradcliffe: I see that. On the other, I feel there's currently high barrier for anybody to just jump in and help — the whole process/setup has many moving parts, missing info, etc
<karmi>
dwradcliffe: Manually massaged the /releases etc, and now finally got there: "FATAL: no pg_hba.conf entry for host "33.33.33.10", user "postgres", database "rubygems_staging", SSL off"
<karmi>
dwradcliffe: I think there's also some mismatch between staging/production — cap tries to run db:migrate with production, but the db is set to _staging, etc
<dwradcliffe>
hmm, that's possible
<dwradcliffe>
I'll have to dig though the code later
<karmi>
dwradcliffe: Also, for switching IPs on vagrant/other vs EC2, I usually just did:
<samkottler>
although that method might have changed since I last used it, which was admittedly a while ago
mootpointer has joined #rubygems
<autojack>
hmm, let me try.
<autojack>
that just gives me true/false
<autojack>
what I'd expect from a method name ending in ?
<autojack>
but there must be a way. from what I've read, the gem CLI tool is just a thin wrapper around the rubygems lib. if that can tell me the versions of stuff, there must be a way to do what I want in code.
<autojack>
I'll look at the Gem api.
<autojack>
if I can find it.
kgrz has quit [Quit: Computer has gone to sleep.]
mootpointer has quit [Quit: Computer has gone to sleep.]
reset has quit [Ping timeout: 272 seconds]
workmad3 has joined #rubygems
dvu has quit [Remote host closed the connection]
fromonesrc has quit [Ping timeout: 256 seconds]
mbull9 has joined #rubygems
dvu has joined #rubygems
fromonesrc has joined #rubygems
mootpointer has joined #rubygems
mootpointer has quit [Max SendQ exceeded]
mootpointer has joined #rubygems
karmi has joined #rubygems
kgrz has joined #rubygems
crandquist has quit [Read error: No route to host]
karmi has quit [Ping timeout: 252 seconds]
mootpointer has quit [Quit: Computer has gone to sleep.]
wedgeV_ has joined #rubygems
wedgeV has quit [Read error: Connection reset by peer]
havenwood has joined #rubygems
<autojack>
I give up. if anyone has a solution, let me know, meanwhile I'll be parsing command line output.