Closed ned21 closed 8 years ago
i'd propose to change it to
# run regularly ccm-fetch. To avoid lost notifications
#
RANDOM_DELAY=45
0 * * * * root (date --iso-8601=seconds --utc; /usr/sbin/ccm-fetch) >>/var/log/ccm-fetch.log 2>&1
random_delay (in minutes) should be supported in all/most cron versions; and it's better than the dangling perl sleep process on all our nodes.
(or did you want a timestamp in on each output line?)
+1 for RANDOM_DELAY, although why 45 and not 60?
On Solaris we ship this via templates rather than pkg. Perhaps we should just move it to templates on Linux too?
nlist("name", "ccm-fetch",
"user", "root",
"timing", nlist("minute", "0", "smear", 59),
"command", "/sbin/ccm-fetch",
ncm-ncd
logs to syslog why this is not the case for ccm-fetch?
If it would then the lack of timestamps on stdout won't be an issue.
because CAF::Reporter
isn't exactly the nicest report/log framework around? the easiest way to have ccm report to syslog is to have it log to a logfile.
any objections against logging to /var/log/ccm-fetch.log
? (and similar for ccm-initialize/purge). or only log to syslog?
@stdweird if that is how we can achieve it then sounds fine to me.
leaves a useful log file in /var/log/ccm-fetch.log but it contains no timestamp. Normally this is not an issue because the timestamp can be inferred from the times in /var/lib/ccm but a timestamp is required when trying to work out why a profile was not downloaded.