Forum: CFEngine Help Subject: Re: copy_from: would like priority list of sources, then silent fail Author: sauer Link to topic: https://cfengine.com/forum/read.php?3,25584,25793#msg-25793
To clarify (or confirm?), this works because the promise key is just the filename, and the first copy_from which works sets the promise as having been kept, so it would not be verified again until the minimum interval had passed? So, subsequent copy attempts would be skipped, due to the promise having already been verified? If so, it seems like there's a race condition here. Say you have cf-agent executing on a 5 minute interval. And you have a five minute elapsed time interval on promises. The first time this runs, file #1 copies succesfully. The second time cf-agent runs, maybe a few of the promises earlier in the policy run slightly quicker than usual (maybe cf-promises doesn't have to run because the inputs are unchanged), so five minutes hasn't elapsed yet when the first element in the copy_from slist is brought up as a candidate for verification. But by the time you get to the third one, the five minutes has elapsed and the promise is again up for validation. The third copy works, leaving your file in an unintended state. Going forward, as long as all of your other promises take exactly the same time to evaluate, maybe the five minute timer and execution time may remain skewed just right so that the promise isn't up for reevaluation on the file1 and file2 checks, but comes up just i n time for file3 to be verified. Your system stays in the wrong state until something happens to bump the phasing between the promise database and the cf-execd timer. While somewhat obscure, it seems like this would be an actual possibility based on the presumption made earlier on, right? I'm partial to defining a set of classes like "file1 exists", "file2 exists&!file1 exists", "file3 exists & !file2 exists" and setting a variable "source" based on those, then grabbing $(source) only if $(source) is defined. It's quite as easy to extend, though. Hopefully someone can explain that my earlier assumption is wrong and that there is no race condition, 'cause Diego's solution is way prettier. :) _______________________________________________ Help-cfengine mailing list Help-cfengine@cfengine.org https://cfengine.org/mailman/listinfo/help-cfengine
