I received a very helpful reply to my questions from Alex van den Bogaerdt <[EMAIL PROTECTED]>
> I would be very surprised if any of your questions have not been > discussed previously. Please refer to the archives. > I'm pretty sure some of the questions (if not all) are answered > (or discussed) during the past few months. > > I do not want to discourage you to ask questions, I just want to > point out that you are traveling a path which others have already > flattened for you. > > Unforunately, the search engine on ethz seems to be down right > now so I cannot point you there. However, using google I found: > http://www.ee.ethz.ch/~slist/rrd-users/msg00817.html > I suggest that you view this article, look at the date, scan the > mail archive around that date for some in depth discussions. I've now read the thread ``RRDtool "GMT-centric" problem?'' and I have a much better understanding of what's going on. It clears up 4 of the 5 questions I posted! Could I make a suggestion? If this had been explained in the man pages (if it is, it's not obvious) -- I would suggest the ``rrdtool'' man page -- it would have saved me *many* hours of confusion. I just noticed that another question was just posted to the mailing list asking about GMT vs. local times. [ As an aside, although rrdtutorial mentions the time values being UTC, it doesn't make clear how the rrd is organized with respect to local or UTC time. Furthermore, the use of the absolute numbers obscures what is going on -- who can make sense out of a plain 9-digit number? ] In addition, the fact that an AT_STYLE time spec is taken to be local time just adds to the confusion. Being able to use --start AT_TIME for rrdcreate makes one believe that one can set the alignment of the time slots. Couldn't the time spec optionally accept "GMT" or "UTC" to avoid the local time conversion? Seems to me that with this, I could just adjust my times in such a way that my local-timed data would be plopped into the correct utc-time time slots. With respect to my question #2 (allowing an AT_STYLE time spec with rrdupdate), is there no reason why this couldn't be done? Having to compute and manipulate the absolute seconds number (and take into account localtime) is a real pain when it's not available directly from a particular data source. To summarize what probably sounds like confused ramblings: if there were a description of how the rrd time slots are aligned to UTC time, and a note that AT-STYLE times are always interpreted as local time, and AT_STYLE time specs (with an optional UTC) were accepted by the rrdupdate command, then you could eliminate almost all of the 9-digit number stuff in the man pages and tutorials. I think this would be a *big* help to folks trying to understand and use rrd's. If I had known about this from reading the man pages, (and rrdupdate had been changed) I would have immediately made my examples: rrdtool create --start '... UTC' ... rrdtool update ... '11:59pm ... [EMAIL PROTECTED]:value:...' and things would have worked as I expected. It might even reduce the number of postings on this mailing list ;-) After all, I suspect that a lot of people create their first test rrd file and feed it localtime day-aligned data to see what's going on -- and are then confused when they read back the results. Hope I've made some sense here... Thanks for your time. Dave Bodenstab -- Unsubscribe mailto:[EMAIL PROTECTED] Help mailto:[EMAIL PROTECTED] Archive http://www.ee.ethz.ch/~slist/rrd-users WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
