Tarantool development patches archive
 help / color / mirror / Atom feed
From: Konstantin Osipov <kostja@tarantool.org>
To: tarantool-patches@freelists.org
Subject: [tarantool-patches] Re: [PATCH 6/6] [RAW] swim: introduce failure detection component
Date: Fri, 29 Mar 2019 21:59:20 +0300	[thread overview]
Message-ID: <20190329185920.GE20712@chai> (raw)
In-Reply-To: <a4660a32b19b398aa71d2aec22b7dbd9de9f1f21.1553078631.git.v.shpilevoy@tarantool.org>

* Vladislav Shpilevoy <v.shpilevoy@tarantool.org> [19/03/20 14:11]:
>  
>  
> +/**
> + * Decode a failure detection message. Schedule acks, process
> + * acks.
> + */
> +static int
> +swim_process_failure_detection(struct swim *swim, const char **pos,
> +			       const char *end, const struct sockaddr_in *src,
> +			       const struct tt_uuid *uuid)
> +{
> +	const char *prefix = "invalid failure detection message:";
> +	struct swim_failure_detection_def def;
> +	struct swim_member_def mdef;
> +	if (swim_failure_detection_def_decode(&def, pos, end, prefix) != 0)
> +		return -1;
> +	say_verbose("SWIM %d: process failure detection's %s", swim_fd(swim),
> +		    swim_fd_msg_type_strs[def.type]);
> +	swim_member_def_create(&mdef);
> +	mdef.addr = *src;
> +	mdef.incarnation = def.incarnation;
> +	mdef.uuid = *uuid;
> +	struct swim_member *member = swim_upsert_member(swim, &mdef);
> +	if (member == NULL)
> +		return -1;
> +	if (def.incarnation < member->incarnation)
> +		return 0;

I think this branch a comment. Why do you not respond to pings
from older incarnations?

> +	if (no_acks_to_gc >= 0)
> +		swim->no_acks_to_gc = no_acks_to_gc;

Why do you actually need a custom no_acks_to_gc? Could you please
add a comment for this variable declaration?

> +enum {
> +	/**
> +	 * Value being used to say that unacknowledged pings
> +	 * should not affect a certain decision about member. For
> +	 * example, regardless of number of unacked pings, never
> +	 * drop a member.
> +	 */
> +	SWIM_NO_ACKS_IGNORE = INT_MAX,
> +};

OK, now I get it. You use it to pin members. I would actually use
bool is_pinned for that. It seems you make no_acks_to_gc a
variable only to use it for pinning. Let's add a separate varaible
for pins.

> +/**
> + * Send a ping to this address. If an ACK is received, the member
> + * will be added.

Could you also please explain in the comment why this is useful?
E.g. why would I want to add a member only after a successful
ping/ack?

> + */
> +int
> +swim_probe_member(struct swim *swim, const char *uri);
> +
> +	SWIM_FD_MSG_TYPE,

The patch is generally OK to push after fixing the minor comments
above. I also changed a few comments on the branch.

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

  reply	other threads:[~2019-03-29 18:59 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-03-20 10:49 [tarantool-patches] [PATCH 0/6] SWIM failure detection draft Vladislav Shpilevoy
2019-03-20 10:49 ` [tarantool-patches] [PATCH 1/6] swim: follow-ups for SWIM anti-entropy Vladislav Shpilevoy
2019-03-29  8:27   ` [tarantool-patches] " Konstantin Osipov
2019-03-29 10:19     ` Vladislav Shpilevoy
2019-03-20 10:49 ` [tarantool-patches] [PATCH 2/6] test: introduce breakpoints for swim's event loop Vladislav Shpilevoy
2019-03-29 18:20   ` [tarantool-patches] " Konstantin Osipov
2019-04-02 12:25     ` Vladislav Shpilevoy
2019-04-02 19:16       ` Vladislav Shpilevoy
2019-04-02 20:40         ` Konstantin Osipov
2019-04-02 21:26           ` Vladislav Shpilevoy
2019-03-20 10:49 ` [tarantool-patches] [PATCH 3/6] test: remove swim_unblock_fd event from swim test harness Vladislav Shpilevoy
2019-03-29 18:22   ` [tarantool-patches] " Konstantin Osipov
2019-04-02 21:26     ` Vladislav Shpilevoy
2019-03-20 10:49 ` [tarantool-patches] [PATCH 4/6] swim: expose enum swim_member_status to public API Vladislav Shpilevoy
2019-03-29 18:24   ` [tarantool-patches] " Konstantin Osipov
2019-04-02 12:25     ` Vladislav Shpilevoy
2019-04-02 13:17       ` Konstantin Osipov
2019-04-02 21:26         ` Vladislav Shpilevoy
2019-03-20 10:49 ` [tarantool-patches] [PATCH 5/6] test: differentiate blocked and closed swim fake fds Vladislav Shpilevoy
2019-03-29 18:25   ` [tarantool-patches] " Konstantin Osipov
2019-04-02 21:26     ` Vladislav Shpilevoy
2019-03-20 10:49 ` [tarantool-patches] [PATCH 6/6] [RAW] swim: introduce failure detection component Vladislav Shpilevoy
2019-03-29 18:59   ` Konstantin Osipov [this message]
2019-04-02 12:25     ` [tarantool-patches] " Vladislav Shpilevoy
2019-04-04 10:20       ` Vladislav Shpilevoy
2019-04-04 12:45       ` Konstantin Osipov
2019-04-04 13:57         ` Vladislav Shpilevoy
2019-04-04 16:14           ` Vladimir Davydov
2019-04-04 16:47             ` Vladislav Shpilevoy
2019-03-27 19:28 ` [tarantool-patches] [PATCH 7/6] swim: make swim_upsert_member returning two values Vladislav Shpilevoy
2019-03-28  8:52   ` [tarantool-patches] " Konstantin Osipov
2019-03-28 11:52     ` Vladislav Shpilevoy

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190329185920.GE20712@chai \
    --to=kostja@tarantool.org \
    --cc=tarantool-patches@freelists.org \
    --subject='[tarantool-patches] Re: [PATCH 6/6] [RAW] swim: introduce failure detection component' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox