So one of the behaviors that seems to wedge puppetmasterd is if you re-initialize a client, the example here being that I installed an older version of puppet on a bunch of machines to test out how things work: * at some point in the past install older rev of puppet - client regiesters itself with the server, etc. - works fine. * recently, remove the old puppet in favor of a new version that has been localized to different directories * fire up puppet and watch puppetmasterd go south At this point I''ve had to: 1) kill the running puppetd client 2) I''ve removed all the legacy dirs on the client node (the fedora core usual places) <- this is probably my bad here, I should have done that to being with to make the system clean 3) remove the old client ssl certificate from the puppetmaster 4) restart puppetmasterd 5) reinitialize the client Then everything normal again. Cheers, Ryan
On May 18, 2007, at 9:25 PM, Ryan Dooley wrote:> So one of the behaviors that seems to wedge puppetmasterd is if you > re-initialize a client, the example here being that I installed an > older > version of puppet on a bunch of machines to test out how things work:I do this constantly to test cert initialization -- I''ve probably had 20 different certs on my laptop -- but I''ve never killed the master before. This is very strange. I even periodically switch to older versions, when I need to check if a given bug is failing in that version.> * fire up puppet and watch puppetmasterd go southWhat does strace say during the process of the master going south? I''ve never seen this kind of problem before. -- No matter how rich you become, how famous or powerful, when you die the size of your funeral will still pretty much depend on the weather. -- Michael Pritchard --------------------------------------------------------------------- Luke Kanies | http://reductivelabs.com | http://madstop.com
>> * fire up puppet and watch puppetmasterd go south > > What does strace say during the process of the master going south? > I''ve never seen this kind of problem before.It''s stuck in select(). The error messages surrounding it are: [2007-05-20 05:00:21] ERROR OpenSSL::SSL::SSLError: /usr/lib/ruby/1.8/openssl/ssl.rb:122:in `accept'' Then dead silence until puppetmasterd is restarted. I''ve a bandaid around it now with monit checking the availability of the 8140/tcp. I need a better test (be nice to have an api for /RPC2/get_server_status or something). If that check fails, monit restarts puppetmasterd. Cheers, Ryan