Forum: CFEngine Help Subject: Re: avoiding double-runs to converge when using template operations Author: jgreer Link to topic: https://cfengine.com/forum/read.php?3,27428,27469#msg-27469
Be aware that all examples and policies produced by Cfengine include the failsafe usage as we've described. That's good, but all of those assume the use of the executor, which is impractical when prototyping. Isn't it much more natural to manually kick off the agent in that scenario? And hasn't that been made harder by abandoning the ability to update policy prior to execution within an agent run? Also, if the agent returns an error, typically from a syntax problem with inputs, it automatically performs a cf-agent -f failsafe.cf thus attempting to correct the issue. So the auto-correct feature is there but you've just missed the new standard operating procedure. I am familiar with that feature, but at least earlier in v3's life (~2 years ago now), I ran into cases in which cf-promises/cf-agent detected syntax errors and silently terminated rather than running the followup failsafe run. I filed bugs at the time, and perhaps all's better now, but since failsafe runs are (for me) cheap and prevent the very scary scenario of mass agent disconnection from the server, it's worth it to me to do an explicit failsafe run as a followup. Finally this new way is more flexible. It allows for the development of self-contained policy files without needing an update.conf. This is especially useful for prototyping. I don't think my proposal precludes the flexibility you describe? If one were allowed the ability to run an arbitrary bundle prior to the standard bundlesequence in a single agent run, such that the standard bundlesequence read policy from $(sys.workdir) into memory, that would allow for that flexibility to be preserved AND for a single invocation of cf-agent to execute on a single revision from version control, where policy and backend managed file revisions might be tightly coupled. That would allow me to trust that a single run of "cf-agent," anytime, on any of my systems, is safe - no need to remember to front it with a failsafe run. That's something we've been able to trust for years, and we'd like to continue to be able to trust it blindly even as we make more use of templatization. _______________________________________________ Help-cfengine mailing list Help-cfengine@cfengine.org https://cfengine.org/mailman/listinfo/help-cfengine
