Hi Klaus, Στις 20 Σεπ 2013, 7:04 μ.μ., ο/η Klaus Aehlig <[email protected]> έγραψε:
> > Hi Spyros, > > thanks for your prompt answer. > >>> - How does your mock-file relate to the text file input? >> >> This file contains reports produced by MonD in contrast to the file of >> the -U option which contains utilization data (cpu, memory, disk and >> network utilization). > > So a different format? OK, that makes sense. We just have to document, > how the behaviour is, if both options are given (hopefully it will be > "use mond data where provided, and defult to the -U data where mond > data is missing"). ack > BTW, your patch series seems to be missing updates > of the htools man pages. The latest update of the htools man pages is the --ignore-dynu option. I can't find anything else. Can you explain it a bit more? > >> If there is a need in the future to query MonDs >> for new Data Collectors it will be possible to override the >> communication to MonDs over the network. >> >>> - Is it possible to keep things backwards compatible, i.e., >>> make querying mond an opt-in feature (e.g., provide some >>> option --mond), rather than an opt-out feature? >> >> I don't think I get that 100% here. By opt-out feature you mean >> --ignore-dynu? Yes, we could provide an opt-in --mond option but how is >> this associated to --ignore-dynu? > > What I mean is: the default behaviour of hbal changes with this > patch series, and I'd like to avoid this, as this might cause all > kinds of problems to people currently using hbal, some even in an > automated way. Therefore my suggestion is to leave the default > behaviour as is to keep backwards-compatibility, and only query > MonD if an appropriate option is given. In this way, people can > leave their scripts and cron-jobs unchaged when upgrading; that's > what I mean by "opt-in": you only get the new behaviour if you > explicitly ask for it (by giving an appropriate option). [Your > implementation is more "opt-out": after an upgrade to the new > ganeti version, you suddently get the new behaviour, unless you > spent some effort to diable it.] > > Concerning the interaction between --mond and --ignore-dynu: why > not the same as that between -U and --ignore-dynu? It works, has > a well defined semantics (the value is read from a file, in order > to be ignored), but probably is not very useful. ack > > I hope, I could make my concerns easier to understand. > > Thanks, > Klaus Thanks, Spyros > > > -- > Klaus Aehlig > Google Germany GmbH, Dienerstr. 12, 80331 Muenchen > Registergericht und -nummer: Hamburg, HRB 86891 > Sitz der Gesellschaft: Hamburg > Geschaeftsfuehrer: Graham Law, Christine Elizabeth Flores
