Thanks Jason for the info. Bear with me, but I'm still unsure why the ENC would run twice, particularly given the dynamic nature of an ENC terminus where it would be connecting to other systems to get information for a catalogue -- those systems are then hit twice, regardless of whether or not the same information is returned, which seems a waste of resources (e.g maybe the external system has a quota limit ?) I see in our logs running twice when the agent, in verbose mode, outputs "Info: retrieving plugin" and then another run when it outputs " Info: Caching catalog for ..."
In relation to applying changes to a virtual hosting setup, I'll look at other options. Given a quick example of say a domain 'enc.example.com' being created, then the customer fails to pay for their domain or hosts elsewhere, such that it should be deleted, the ENC should then provide this account for deletion in the catalogue, so that it is removed on the relevant server. I understand that this is changing the catalogue information across ENC runs but how else would this change be provided to the agent ? Regards James On Wednesday, 15 January 2014 12:40:39 UTC+11, Jason Antman wrote: > > James, > > I vaguely remember seeing this 'node_terminus called twice' thing in the > past. The simple answer, though I know it's not what people want to hear, > is that the ENC should always return the right information. If you want to > modify the content of the catalog based on something that happens between > runs (which, by the way, I would highly suggest against, and suggest that > if you're doing that, something in your puppet configuration is amiss), you > should be checking somewhere like PuppetDB, not changing it based on how > many times the ENC script is run. Aside from the problem you're having now, > what happens if there's a timeout, or some other failure after the ENC > script is called but before the catalog is applied? > > -Jason > > On 01/14/2014 06:20 PM, James Ellis wrote: > > Hi, chanced across this discussion when I noticed an ENC was being > called twice. I understand I may not be using the ENC terminus exactly as > it's been designed, but it's unexpected that it was called twice. Also > worth noting that I can't see a note about the ENC being called twice here: > http://docs.puppetlabs.com/guides/external_nodes.html > > In my case, I'm using an ENC to push virtual host changes to an agent > running a web server, the YAML returned by the ENC uses create_resources to > dynamically add resources to the catalogue. > I observed via logging in the ENC script that on the first run, the ENC > was excecuted but the catalogue was not applied, on the second run the > catalogue was applied on the agent. > This causes problems where we use an API to dynamically apply resources > to a catalogue (1st run gets the catalogue resources, returns 'OK' to the > API, 2nd run then tries to get resources but gets nothing as the 'OK' sent > to the API has effectively modified the resources to be applied). > > I've worked around this, for now, by using a lock file, so that the 'OK' > API call is only run once but this still applies two calls to the API to > dynamically get resources for the catalogue, where only one is required. I > double checked the master and the agent configs, and the master only shows > the ENC being referenced once and there is one agent being run, only. > > Based on this, is there any way the agent can be set to call the ENC > once only ? The only argument to the script is the agent hostname and there > is no apparent difference in the environment of the first and second ENC > calls. > > Using 3.4 O/S on ubuntu with the following agent command (run as root > manually to debug) : > puppet agent --no-usecacheonfailure --onetime --no-daemonize --server > valid.server --verbose > > > Thanks > James > > > On Monday, 23 September 2013 23:59:45 UTC+10, jcbollinger wrote: >> >> >> >> On Friday, September 20, 2013 12:05:17 PM UTC-5, Greg Sutcliffe wrote: >>> >>> Is this puppet3? As I recall, in puppet3, the master makes a separate >>> call to the enc to determine the environment the should authoritatively be >>> in. Once that's established, it makes a second call to get the classes and >>> parameters. >>> >> >> Not exactly, but that may well be the right track. It would be pointless >> for the master to run the ENC more than once for catalog compilation, for >> it would have no reason to expect that the ENC's output would change. >> HOWEVER, the master's file server may need to run the ENC again to >> determine the environment from which to serve 'source'd files. >> >> >> John >> >> -- > You received this message because you are subscribed to the Google Groups > "Puppet Users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to puppet-users...@googlegroups.com <javascript:>. > To view this discussion on the web visit > https://groups.google.com/d/msgid/puppet-users/373455ff-43f4-4f27-8667-ab896c240ea6%40googlegroups.com > . > For more options, visit https://groups.google.com/groups/opt_out. > > > -- You received this message because you are subscribed to the Google Groups "Puppet Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to puppet-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/puppet-users/0c0d3a2f-598d-4f84-a8b2-5beaf67d2820%40googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
