The watchdog.conf man page of watchdog-5.2.4_p5 writes about the logtick option: "logtick = <logtick> If you enable verbose logging, a message is written into the syslog or a logfile. While this is nice, it is not necessary to get a message every 10 seconds which really fills up disk and needs CPU. logtick allows adjustment of the number of intervals skipped before a log message is written. [...]" This sounds very useful but there are two problems with its implementation: 1. The logtick option only applies for the "I'm alive" message in the main loop ("[watchdog] still alive after x seconds = y interval(s)"). All other messages are written every "interval" seconds and fill up the logfiles anyway (eg. "[watchdog] current load is 0 0 0"). For the reason see pidfile.c for example. It just checks: if (verbose) syslog(LOG_INFO, "was able to ping process ... 2. The main loop prints the "I'm alive" message only once because of an programming error: See watchdog.c. The check to run the syslog() call is (--ticker == 0) but if verbose is true the value of ticker will be set to 0. As ticker is set nowhere else the check (--ticker == 0) will never be true again. --> if(logtick && (--ticker == 0)) { ticker = logtick; if (verbose) { count += logtick; --> ticker = 0; syslog(LOG_INFO, "still alive after %ld seconds = ... } }
Please report this issue upstream.
I reported this and all other watchdog issues to upstream some month ago but no reaction so far. Just FYI.
(In reply to comment #2) > I reported this and all other watchdog issues to upstream some month ago but no > reaction so far. Just FYI. Oh, I thought your proposed changes was included in 5.2.5_p1?
(In reply to comment #3) > (In reply to comment #2) > > I reported this and all other watchdog issues to upstream some month ago but no > > reaction so far. Just FYI. > > Oh, I thought your proposed changes was included in 5.2.5_p1? > Just for the bug searchers... --logtick option works for me in 5.2.5_p1.
