Thanks, Mark and Seva. If you produce a beta RPM (RH-5/Centos-5) with this, let me know and I can try it out for you.
Going more theoretical, this raises a worrying question. I had written the bundle that contained the "delete_lines" and "insert_lines" on the assumption that this pair of operations, taken together, would have promised to deliver an atomic "all or nothing" transaction. Is this not the case? If not, then by what means can one achieve an atomic set of actions? I realise that fully generalising this might be very tricky, perhaps even verging on impossible. But equally there are probably cases (e.g. a set of file edits in a single file) where it is both highly desirable and realistically achievable. And there are probably similar cases (e.g. setting up network interfaces on some OSes) where a set of edits across more than one file, as an atomic action, may also be highly desirable. (Indeed, my phrase "highly desirable" might, in some circumstances, need to be read as "essential".) Anyway, thanks for the attention and replies. -- David Lee Mark Burgess wrote: > Thanks for reminding me about this. I was sure this had already been done. > > On 03/14/2011 09:05 PM, Seva Gluschenko wrote: >> David, >> >> if by occasion you experience concurrent cf-agent runs, there's a >> situation possible when one instance locks insert_lines promises and >> another one locks delete_lines promises. There were discussions with >> Mark about locking the whole file to prevent such undesired >> concurrency overlaps, and I hope he will implement it in the next >> release. >>[...] _______________________________________________ Help-cfengine mailing list Help-cfengine@cfengine.org https://cfengine.org/mailman/listinfo/help-cfengine
