[tarantool-patches] Re: [PATCH v2 1/2] swim: disseminate event for log(cluster_size) steps
Vladislav Shpilevoy
v.shpilevoy at tarantool.org
Thu Jul 4 02:30:51 MSK 2019
Hi! Thanks for the review!
>> diff --git a/src/lib/swim/swim.c b/src/lib/swim/swim.c
>> index bb9e9f519..9647094f5 100644
>> --- a/src/lib/swim/swim.c
>> +++ b/src/lib/swim/swim.c
>> @@ -616,7 +616,24 @@ swim_register_event(struct swim *swim, struct swim_member *member)
>> rlist_add_tail_entry(&swim->dissemination_queue, member,
>> in_dissemination_queue);
>> }
>> - member->status_ttd = mh_size(swim->members);
>> + /*
>> + * Logarithm is a perfect number of disseminations of an
>> + * event.
>> + *
>> + * Firstly, it matches the dissemination speed.
>> + *
>> + * Secondly, bigger number of disseminations (for example,
>> + * linear) causes events and anti-entropy starvation in
>> + * big clusters, when lots of events occupy the whole UDP
>> + * packet, and factually the same packet content is being
>> + * sent for quite a long time. No randomness. Anti-entropy
>> + * does not get a chance to disseminate something new and
>> + * random. Bigger orders are redundant and harmful.
>> + *
>> + * Thirdly, logarithm is proved by the original
>> + * SWIM paper as the best option.
>> + */
>> + member->status_ttd = ceil(log2(mh_size(swim->members)));
I've changed this place to
member->status_ttd = ceil(log2(mh_size(swim->members))) + 1;
It allows to do not break the tests in this commit, and fixes a bug,
when status_ttd became negative for 'self'. Because ceil(log2(1)) = 0,
and on a next round step it became -1, -2, etc.
I didn't push the patch yet, if you have anything against that.
More information about the Tarantool-patches
mailing list