load graph displaying incorrect time
I have Membase installed on CentOS 6, but the bucket "load graph" is displaying the time incorrectly. I realize this probably has something to do with the underlying OS settings, but the "date" command for the OS shows correctly so I'm lost as to what it might be.
What do I need to check/reset/fiddle with in order to get the Membase graphs to display the correct time?
Thanks!
As far as I know, it has always been incorrect. It always displays 5 hours earlier than what the actual local time is. It sounds like a timezone issue, but I changed the timezone, and the graph just shifts accordingly. It's almost like membase is reading the system time as UTC, then adjusting for my timezone (EST), even though the system time is set up as EST.
hi.
In order to diagnose your case can you please give us the following information:
output of:
wget -O- --progress-type=dot --user=Administrator --password= --post-data='{erlang:now(), os:timestamp(), os:cmd("date -u")}.' http://:8092/diag/eval | cat
as well as what's timezone on OS that runs browser ? Does it differ from timezone on erlang node?
Output:
{{1324,285977,548787},{1324,303576,902067},"Mon Dec 19 14:06:16 UTC 2011\n"}
The timezones on the machine I am viewing/browsing the console from and the server both have the same timezone settings (EST).
Thanks for info. Somehow erlang:now() and os:timestamp() give almost 5 hours of difference. That's very weird and is likely the difference you're seeing.
I need to take a look at erlang VM source to see how this could happen. Interestingly, on my box both functions call same libc function - gettimeofday.
I've filed bug for your case: http://www.couchbase.org/issues/browse/MB-4554
BTW, what's uptime of that membase node ?
No problem, thanks for the help. :)
The install itself is about 3 weeks old, the actual server uptime is just over 3 days now (due to a server RAM upgrade).
Did this happen over the course of time, or has it always been out of synch? Has your clock jumped quite a bit?
Also, do you have multiple nodes in the cluster? We occasionally have console issues if there is a large drift.