cumin r2751 The stats graph for a schedd coming from north-15 is drawing off the right hand side of the graph. Theory is this is an issue with north-15's clock. However, if true, such an issue would be a nightmare to debug by a customer, so some means to handle clock skew should be implemented.
north-15 was sending stats with a future time (relative to the machine cumin is running on). Solution implemented: If the recTime sent in the update message is in the future, use the current time instead.
See https://bugzilla.redhat.com/show_bug.cgi?id=470315 because of which I can not access any shedulers and see the graphs on rev 2751. But nevertheless, I have reproduced the bug on rev 3002: 0. turn off NTP, set time 20 mins back on local machine running cumin (via 'date' command) 1. clearly install cumin, including database init 2. log in, add north-15.lab.bos as a broker 3. Grid -> Schedulers -> <one_scheduler> Maybe it is not the same as when it continuously started at (as I understand it) non-visible right-hand side, but the graphing when the time is not in-sync is definitely strange in rev 3002. See the attached picture - for at least 5 minutes, there was no line and suddenly it appeared for the alredy passed time. And then, when I refreshed the page 5 more minutes later, the weakly-colored-block moved to right and was not continuous (still the same length); all repeating. Kind of: ------------oooooooooooooooo--| --------oooooooooooooooo------| ----oooooooooooooooo----------| oooooooooooooooo--------------| ------------oooooooooooooooo--| ... and so on and so forth ...
Created attachment 327084 [details] The moving non-continuous bars (on both graphs). Here is cropped screenshot.
I think this is explained by this defect: The charts only draw segments for data points on the graph. 10 mins 0 mins o | o----------o | A B C Since point A is off the end, no A-B segment is drawn. I think we should try to address this in 1.1.1.
Addressed in revision 3046. I'm querying 1.5 x period seconds, and then bailing out after I get the first sample outside the period window.
Wonderful! Works as I expect it to. Verified on RHEL5.3 i386, cumin-0.1.3073-1.el5 Thanks, Jasan
I am considering the original time-not-in-sync issue solved as well and much earlier as there was no off-graph drawing already in rev 3002.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHEA-2009-0434.html