Every repository with this icon (
Every repository with this icon (
| Description: | Run a self-healing, auto-scaled and monitored cloud simply, in the clouds, on nearly any hardware, such as EC2, eucalyptus and vmware edit |
-
this issue is a placeholder for notes on using centos and more restrictive environments.
Comments
-
i think its pulling from rubyforge? specify source more explicitly per gem in the repo.
Comments
-
2 comments Created 2 months ago by tralfamadudeUsing an alternative port for ssh should be supportedauserxIf a non-standard port for sshd listening (or other ssh, scp options) on spawned nodes is used, the following construction, if supplied at the beginning of a clouds.rb file, is supposed to override the default port (or other ssh or scp -o settings):
module CloudProviders
module Connectionsdef ssh_options(opts={}) o = {"-p" => '1999', "-i" => keypair.full_filepath, "-o" =>"StrictHostKeyChecking=no" }.merge(opts) o.collect{ |k,v| "#{k} #{v}"}.join(' ') end def scp_options(opts={}) o = {"-P" => '1999', "-i" => keypair.full_filepath, "-o" =>"StrictHostKeyChecking=no" }.merge(opts) o.collect{ |k,v| "#{k} #{v}"}.join(' ') endend end
However, these module attributes are not used uniformly throughout PoolParty.
Comments
In your using block, try setting this as options:
using :ec2, :"-P" => "1999" do end
tralfamadude
Thu Sep 17 16:30:15 -0700 2009
| link
Workaround: if your provisioner machine allows, setting your ssh defaults (usually ssh_config under /etc or /etc/ssh/) to have "Port 1999" will cause the right thing to happen.
A related, interesting fact: sshd can listen on more than one port, so standard port 22 can be opened temporarily.
-
A verbose configure trace should list all the "def" steps performed by name (like "performing Foo.baz on host H" or "performing Foo.bar on prototype image for cloud Foop" or "performing Foo.fum on provisioner host).
The basic idea is that looking at verbose output should give amble clues as to what def section of which rb source file:
is being performed
is executing on what machine
is probably emitting subsequent verbose logging messages
is affecting which hosts
The would go a long way for troubleshooting and debugging.
Comments
-
guarantees about state of clouds after a command returns
0 comments Created 2 months ago by tralfamadudeThere should be guarantees about state of clouds after a command returns (are all commands synchronous? Configuration seems to be, but cloud terminate is not; possibly have a command line flag to wait until done). This enables scripting.
I noticed this after I followed a "cloud terminate" quickly with a ec2-revoke and ec2-delete-group commands that failed because termination was not yet completed.
Comments
-
Running poolparty 1.3.16:
cloud-list provides listing- correctly.
Attempting to to run 'cloud-start' throws:
warning: peer certificate won't be verified in this SSL session
/var/lib/gems/1.8/gems/auser-poolparty-1.3.16/bin/../lib/keypair.rb:30:inexpand_path': can't convert nil into String (TypeError) from /var/lib/gems/1.8/gems/auser-poolparty-1.3.16/bin/../lib/keypair.rb:30:infull_filepath'
from /var/lib/gems/1.8/gems/auser-poolparty-1.3.16/bin/../lib/keypair.rb:19:in `exists?' from /var/lib/gems/1.8/gems/auser-poolparty-1.3.16/bin/../lib/keypair.rb:90:in `keypair_found?' from /var/lib/gems/1.8/gems/auser-poolparty-1.3.16/bin/../lib/keypair.rb:73:in `send' from /var/lib/gems/1.8/gems/auser-poolparty-1.3.16/bin/../lib/keypair.rb:73:in `valid?...and more which I can provide.
Thanks-
MK
Comments
portertech
Sat Nov 07 17:16:10 -0800 2009
| link
Still an issue for me. V 1.3.17 ../lib/keypair.rb:32:in `expand_path': can't convert nil into String (TypeError)
The gem version should be 1.4. Grab it off gemcutter.org
gem sources -a http://gemcutter.org gem install poolparty -
security_group does not use the access_key and secret_access_key given to the using block
1 comment Created about 1 month ago by contentfreeGiven:
using :ec2 do access_key 'something...' secret_access_key 'mysecret...' security_group do ...security_group fails with :access_key_id not given. It's trying to instantiate a new instance of EC2 instead of using the instance from the using block. If it's going to make a new instance it should at least use the keys given to the containing block. This problem is easy to reproduce if you do not set ENV['EC2_ACCESS_KEY'] in your .profile or on the command line.
Comments
contentfree
Mon Oct 26 21:19:47 -0700 2009
| link
I believe this is fixed. Right?
-
Race condition when starting a cloud?
0 comments Created about 1 month ago by contentfreeIt seems more often than not when I start a cloud and its security group doesn't yet exist and will be created by poolparty, cloud-start will fail (supposedly after the security group has been created) when connections to port 22 get refused.
I'm able to relatively reliably reproduce the issue by starting from a virgin state: No instances, a custom security group expressed in my clouds.rb and no security groups created.
Comments
-
The path given to chef_repo needs to be maintained when creating the solo.rb file
0 comments Created about 1 month ago by contentfreeCurrently, the paths in solo.rb are hardcoded to /etc/chef/chef-repo/site-cookbooks and /etc/chef/chef-repo/cookbooks. The problem is, when you call chef_repo in clouds.rb it simply copies the directory directly over so
chef_repo "chef"would create /etc/chef/chef/* which doesn't match the expectations in the hardcoded solo.rb file.Comments
-
Paths given to chef_repo aren't expanded relative to where cloud-* is being run
1 comment Created about 1 month ago by contentfreeFor example, say you have
chef_repo "chef"in your clouds.rb in a hypothetical "config" directory and run cloud-start from the parent directory giving the command line switch "-c config/clouds.rb".chef_repowill not find the "chef" directory so it will not copy your cookbooks.The expectation is that it would find the directory relative to the clouds.rb file.
Comments
contentfree
Mon Oct 26 21:19:04 -0700 2009
| link
Any movement on this?
- EC2▾
- Immediate▾
- Not pressing▾
- auser▾
- binaries▾
- bug▾
- doc▾
- fairchild▾
- feature request▾
- jashmenn▾
- resolvers▾
- vmware▾
- website▾
- Apply to Selection
-
Change Color…
Preview:preview
- Rename…
- Delete











in my build_centos.sh I have made the follwoing change to set the hostname:
Replaced this block:
Set a hostname
if [ -z
hostname]; thenhostname "poolparty" echo poolparty > /etc/hostname /etc/init.d/hostname.sh start 2>/dev/null fi
With:
Set a hostname
Grab the public IP address....
PUBLIC_IP=
/usr/local/bin/ec2-metadata -o|cut -d " " -f 2echo $PUBLIC_IP>/tmp/ip_addr
=
hostname|cut -d "-" -f 1if [ $HOST_NOT_SET = "domU" ]; thenhostname $HOSTNAME sed -i.bak s/localhost.localdomain/$HOSTNAME/g /etc/sysconfig/network echo "127.0.0.1 $HOSTNAME localhost.localdomain localhost" >/etc/hosts
fi
While I have ec2-metadata installed on my images, you can also use the
"curl http://169.254.169.254/latest/meta-data/local-ipv4" method
The main problem is this is an instance by instance setup. I am still trying to figure out how to do a cloud-start, and ether explicitly assign from an array possibly, or use a pattern (i.e. cloud-start 5 instances, with the hostnames webX, webX+1, etc)