In riak_repl_wm_stats a UTC datetime is passed into httpd_util:rfc1123/1, which expects a local date. It causes the stats to blow up when clocks go forward for DST. rfc1123 takes a local datetime and throws an exception when you pass a datetime that sits within the mystical non-existent hour when moving clocks forward. It should be handled gracefully, and there's a patch in OTP 20.1 for it. But we should be passing in a local time anyway. The net result is that we lose the ability to monitor for an hour during this time period.
(riak@127.0.0.1)1> Seconds = 63657709200. %% riak_core_util:moment() to produce current UTC seconds
63657709200
(riak@127.0.0.1)2> Date = calendar:gregorian_seconds_to_datetime(Seconds).
2017,3,26},{1,0,0
(riak@127.0.0.1)3> Formatted = httpd_util:rfc1123_date(Date).
exception error: no case clause matching []
in function httpd_util:rfc1123_date/1 (httpd_util.erl, line 344)
Fixed by converting to a local datetime with calendar:universal_time_to_local_time/1:
(riak@127.0.0.1)13> Seconds = 63657709200. %% riak_core_util:moment() to produce current UTC seconds
63657709200
(riak@127.0.0.1)14> Date = calendar:gregorian_seconds_to_datetime(Seconds).
2017,3,26,1,0,0
(riak@127.0.0.1)15> DateLocal = calendar:universal_time_to_local_time(Date).
2017,3,26,2,0,0
(riak@127.0.0.1)16> Formatted = httpd_util:rfc1123_date(DateLocal).
"Sun, 26 Mar 2017 01:00:00 GMT"
Think this is the same as described in basho/riak_core#773.
In
riak_repl_wm_statsa UTC datetime is passed intohttpd_util:rfc1123/1, which expects a local date. It causes the stats to blow up when clocks go forward for DST. rfc1123 takes a local datetime and throws an exception when you pass a datetime that sits within the mystical non-existent hour when moving clocks forward. It should be handled gracefully, and there's a patch in OTP 20.1 for it. But we should be passing in a local time anyway. The net result is that we lose the ability to monitor for an hour during this time period.Fixed by converting to a local datetime with
calendar:universal_time_to_local_time/1:Think this is the same as described in basho/riak_core#773.