[tarantool-patches] Re: [PATCH 6/6] swim: introduce payload

[Konstantin Osipov]

* Vladislav Shpilevoy <v.shpilevoy at tarantool.org> [19/04/12 01:25]:
> Payload is arbitrary user data disseminated over the cluster
> along with other member attributes.

Please see my comments inline.

> Part of #3234
> ---
>  src/lib/swim/swim.c         | 155 ++++++++++++++++++++++++++--
>  src/lib/swim/swim.h         |   8 ++
>  src/lib/swim/swim_proto.c   |  31 +++++-
>  src/lib/swim/swim_proto.h   |  41 +++++++-
>  test/unit/swim.c            | 195 +++++++++++++++++++++++++++++++++++-
>  test/unit/swim.result       |  32 +++++-
>  test/unit/swim_test_utils.c |  62 ++++++++++++
>  test/unit/swim_test_utils.h |  18 ++++
>  8 files changed, 525 insertions(+), 17 deletions(-)
> 
> diff --git a/src/lib/swim/swim.c b/src/lib/swim/swim.c
> index 2dac6eedd..2be1c846b 100644
> --- a/src/lib/swim/swim.c
> +++ b/src/lib/swim/swim.c
> @@ -308,6 +308,38 @@ struct swim_member {
>  	 * allow other members to learn the dead status.
>  	 */
>  	int status_ttl;
> +	/** Arbitrary user data, disseminated on each change. */
> +	char *payload;
> +	/** Payload size, in bytes. */
> +	uint16_t payload_size;
> +	/**
> +	 * True, if the payload is thought to be of the most
> +	 * actual version. In such a case it can be disseminated
> +	 * farther. Otherwise @a payload is suspected to be

farther -> further.

> +	 * outdated and can be updated in two cases only:
> +	 *
> +	 * 1) when it is received with a bigger incarnation from
> +	 *    anywhere;
> +	 *
> +	 * 2) when it is received with the same incarnation, but
> +	 *    local payload is outdated.
> +	 *
> +	 * A payload can become outdated, if anyhow a new
> +	 * incarnation of the member has been learned, but not a
> +	 * new payload.

Please describe how this can happen: the payload message is lost,
but the server with the new payload responeded to ping request, 
and a new incarnation arrived along with the ack.

> In such a case it can't be said exactly

Nit: in such a case is very rarely used. Why not simply: in this case?

> +	 * whether the member has updated payload, or another
> +	 * attribute. The only way here is to wait until the most
> +	 * actual payload will be received from another instance.
> +	 * Note, that such an instance always exists - the payload
> +	 * originator instance.
> +	 */
> +	bool is_payload_up_to_date;
> +	/**
> +	 * TTL of payload. At most this number of times payload is
> +	 * sent as a part of dissemination component. Reset on
> +	 * each payload update.
> +	 */
> +	int payload_ttl;

As agreed, let's rename ttl to ttd across the board and update the
comments to say that ttd is time to disseminate.

> +/** Update member's payload, register a corresponding event. */
> +static inline int
> +swim_update_member_payload(struct swim *swim, struct swim_member *member,
> +			   const char *payload, uint16_t payload_size,
> +			   int incarnation_increment)

Passing incarnation_increment raises a lot of questions when
reading the code. I would not use this function from the
constructor - I don't see any issue in copy-pasting 3 lines of
code into the constructor to make both branches simpler.

Or I would ban passing payload to the constructor and make this
function public.

>  static int
>  swim_encode_member(struct swim_packet *packet, struct swim_member *m,
> -		   struct swim_passport_bin *passport)
> +		   struct swim_passport_bin *passport,
> +		   struct swim_member_payload_bin *payload_header,
> +		   bool is_payload_needed)

Please rename is_payload_needed  -> encode_payload. There is no
comment for encode_payload and how it is used.
I don't see why you can't move all the decision making about
whether to encode the payload or not into this function.

The payload should be encoded if:
- its not null
- payload ttl is greater than zero
- it's not trustworthy. (is_payload_up_to_date). I would btw
  rename is_payload_up_to_date to is_payload_trustworthy - this
  name would be closer to truth.

Why can't you look at these conditions once in swim_encode_member
rather than evaluate them outside this function?

BTW, there is no harm in encoding an empty payload all the time
(e.g. mp_bin of size 0). This would make the code simpler - you
would only need to look at payload_size to see if payload exists.

> +int
> +swim_set_payload(struct swim *swim, const char *payload, uint16_t payload_size)
> +{
> +	if (payload_size > MAX_PAYLOAD_SIZE) {
> +		diag_set(IllegalParams, "Payload should be <= %d",
> +			 MAX_PAYLOAD_SIZE);
> +		return -1;
> +	}
> +	return swim_update_member_payload(swim, swim->self, payload,
> +					  payload_size, 1);
> +}

Like I said above, if there is swim_set_paylaod, there is no need
to supply payload in swim_new_member(). 

> +	def->payload_size = -1;

Ugh.


-- 
Konstantin Osipov, Moscow, Russia, +7 903 626 22 32
http://tarantool.io - www.twitter.com/kostja_osipov