Fullsync starts at unexpected interval because of a couple of timers running.

[ksauzz]

It seems a couple of fullsync timers are running if fullsync_on_connect is true and fullsync_interval is set. So fullsync starts at unexpected intervals.

ZD #5638, in Japanese, but graphs might be very helpful

to reproduce

1. set fullsync_interval
2. make stagedevrel
3. run the script to setup MDC version 3.

it didn't need to execute riak-repl fullsync start <cluster-name>

• source cluster: dev1, dev2, dev3
• site cluster: dev4, dev5, dev6

console.log to indicate a scheduled time at source cluster (fullsync_on_connect is enable)

• fullsync_interval=2min
• fullsync_on_connect=true

console.log to indicate a scheduled time at source cluster (fullsync_on_connect is disable)

• fullsync_interval=2min
• fullsync_on_connect=false

[bookshelfdave]

Can you try to build riak_ee 1.4 from source and see if this branch of riak_repl helps?
https://github.com/basho/riak_repl/tree/dip_schedule_fix

git clone git@github.com:/basho/riak_ee.git
cd riak_ee
git checkout 1.4
./rebar get-deps
cd deps/riak-repl
git checkout dip_schedule_fix

If a new timer is scheduled while an old timer is running, the old timer is now canceled. This can be seen in the logs: "Fullsync schedule timer for ~p reset"

[ksauzz]

@metadave Great!

I built and ran riak_repl of dip_schedule_fix on stagedevrel. A new timer works fine. Thanks.

• fullsync_interval=2min
• fullsync_on_connect=true

[cmeiklejohn]

Should this be closed? Going to move to 2.1 until we confirm.

[ksauzz]

Unfortunately, no. At least, Riak 1.4.8 has this issue. And I couldn't test on Riak 2.0.0beta1 because it looks fullsync timer never trigger fullsync. I'll file new ticket about it.

[ksauzz]

It's my fault. The timer triggers fullsync well on 2.0.0beta1 but a couple of timers still run on it.

[shino]

Squashed and rebased Dave's working branch onto dbfddec (which has the tag 1.4.10 and commit message "Update deps for 1.4.10")

Compare view: https://github.com/basho/riak_repl/compare/1.4...dip_schedule_fix-rebased?expand=1