3.0 config: update replication tutorials by andreyaksenov · Pull Request #3862 · tarantool/doc

andreyaksenov · 2023-11-17T13:29:05Z

Created 3 new tutorials for each failover mode:

The Configuring synchronous replication section is removed as it demonstrates not the best practices. Information about enabling sync replication added in the main topic and API docs:

Managing leader elections is moved to Concepts as it doesn't fit into the Tutorials section:

Replication (index page)
Managing leader elections (moved from tutorials)

Added information about a new box.info.election.leader field:

box.info.election

Mentioned new tutorials in configuration reference for the replication.failover option:

replication.failover

Totktonada

I saw several drafts of this patchset and it was generally OK for me. I'm going to approve without looking over one more time.

I would only note that it is easy to become entangled with all these replication reconfiguration steps. I would illustrate them with some pictures, if possible (of course, it is not a blocker for this pull request).

p7nov

I've checked:

They're fine by me, just some minor improvement thoughts.

p7nov · 2023-12-05T10:56:14Z

-    replication source(s), and
-*   :ref:`read_only <cfg_basic-read_only>` which is ``true`` for a
-    replica and ``false`` for a master.
+Prerequisites


I feel a lack of some general description of what we'll do. Something like "We'll start a cluster with a manual failover, check how the replication works, and switch master manually."

Totally forgot about intros, added.

p7nov · 2023-12-05T10:58:24Z

+Reloading configuration
+~~~~~~~~~~~~~~~~~~~~~~~
+
+After adding ``instance003`` to the configuration and starting it, configurations on all instances should be reloaded to allow ``instance001`` and ``instance002`` to get data from the new instance in case it becomes a master:


This is exactly the case where passive voice confuses readers :)
I couldn't understand by whom the configuration should be reloaded until I read the step 2 below.

Thanks, will fix :)

p7nov · 2023-12-05T11:00:07Z

+Removing an instance from the configuration
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+1.  Remove ``instance003`` from the ``instances.yml`` file:


Minor: it would be natural to remove instance001, since we've already took leadership from it.

p7nov · 2023-12-05T11:00:47Z

+
+..  _replication-automated-failover-tt-env:
+
+Prerequisites


Same about the description.

p7nov · 2023-12-05T11:01:24Z

+        ...
+
+4.  Execute ``box.info.replication`` to check a replica set status.
+    Make sure that ``upstream.status`` and ``downstream.status`` are ``follow`` for ``instance002`` and ``instance003``.


Suggested change

Make sure that ``upstream.status`` and ``downstream.status`` are ``follow`` for ``instance002`` and ``instance003``.

Make sure that ``upstream.status`` and ``downstream.status`` are ``follow`` for ``instance001`` and ``instance003``.

Make sure that upstream.status and downstream.status are follow for instance002 and instance003.

Looks like the old description is correct, returned back:

p7nov · 2023-12-05T11:01:45Z

+Adding data
+~~~~~~~~~~~
+
+To check that replicas (``instance001`` and ``instance003``) get all updates from the master(``instance002``), follow the steps below:


Suggested change

To check that replicas (``instance001`` and ``instance003``) get all updates from the master(``instance002``), follow the steps below:

To check that replicas (``instance001`` and ``instance003``) get all updates from the master (``instance002``), follow the steps below:

p7nov · 2023-12-05T11:04:07Z

+
+3.  Use the ``select`` operation on ``instance001`` and ``instance003`` to make sure data is replicated.
+
+4.  Check that the 1-st component of :ref:`box.info.vclock <box_introspection-box_info>` values are the same on all instances:


1-st looks a bit weird.
I get that it's not the first but corresponding to key 1. Maybe call it "1 component" ("1" digit in code style)?

Agree, took this from the old changelog. Will fix.

p7nov · 2023-12-05T11:05:15Z

+Choosing a leader manually
+--------------------------
+
+1.  Make sure that :ref:`box.info.vclock <box_introspection-box_info>` values (excluding the 0-th components) are the same on all instances:


Maybe excluding > except?

xuniq · 2023-12-06T12:11:21Z

+Leader election doesn't work correctly if the election quorum is set to less or equal
+than ``<cluster size> / 2`` because in that case, a split vote can lead to
+a state when two leaders are elected at once.


Suggested change

Leader election doesn't work correctly if the election quorum is set to less or equal

than ``<cluster size> / 2`` because in that case, a split vote can lead to

a state when two leaders are elected at once.

Leader election doesn't work correctly if the election quorum is set to less or equal

than ``<cluster size> / 2``. In that case, a split vote can lead to

a state when two leaders are elected at once.

xuniq · 2023-12-06T12:24:26Z

+Step 1: Configuring a failover mode
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+First, set the :ref:`replication.failover <configuration_reference_replication_failover>` option to ``manual``:


Suggested change

First, set the :ref:`replication.failover <configuration_reference_replication_failover>` option to ``manual``:

Set the :ref:`replication.failover <configuration_reference_replication_failover>` option to ``manual``:

xuniq · 2023-12-06T15:07:10Z

+            ...
+
+3.  Execute ``box.info.replication`` to check a replica set status.
+    For ``instance002``, ``upstream.status`` and ``downstream.status`` should be ``follow``.


Suggested change

For ``instance002``, ``upstream.status`` and ``downstream.status`` should be ``follow``.

For ``instance001``, ``upstream.status`` and ``downstream.status`` should be ``follow``.

Looks like the current description is correct:

This was linked to issues Nov 17, 2023

[Config] How-to: replicaset configuration #3654

Closed

[Config] How-to: Master-master configuration #3655

Closed

[Config] How-to: failover using Raft (failover: election) #3656

Closed

Document box.info.election #3680

Closed

andreyaksenov force-pushed the 3.0-config-replication-tutorials branch 6 times, most recently from 6994648 to d230e43 Compare November 20, 2023 14:56

andreyaksenov force-pushed the 3.0 branch from 2d5e760 to 68f8837 Compare November 21, 2023 08:09

andreyaksenov force-pushed the 3.0-config-replication-tutorials branch 19 times, most recently from 7153397 to 2355b9c Compare November 23, 2023 12:25

andreyaksenov force-pushed the 3.0-config-replication-tutorials branch 9 times, most recently from e15a38f to 2b08f1d Compare December 1, 2023 08:50

andreyaksenov marked this pull request as ready for review December 5, 2023 06:53

andreyaksenov force-pushed the 3.0-config-replication-tutorials branch 2 times, most recently from 2c8e3b6 to 8aedebc Compare December 5, 2023 07:25

andreyaksenov requested review from Totktonada, p7nov and xuniq December 5, 2023 07:40

Totktonada approved these changes Dec 5, 2023

View reviewed changes

p7nov approved these changes Dec 5, 2023

View reviewed changes

andreyaksenov force-pushed the 3.0-config-replication-tutorials branch 2 times, most recently from b189d5f to 9af33f9 Compare December 5, 2023 12:07

andreyaksenov force-pushed the 3.0 branch from 7441be9 to 61f563c Compare December 6, 2023 10:06

xuniq approved these changes Dec 6, 2023

View reviewed changes

andreyaksenov added 6 commits December 7, 2023 10:10

3.0 configuration: update replication tutorials

6232707

3.0 configuration: update master-master sample

db417ee

3.0 configuration: update replication tutorials (review fixes 1)

c35b016

3.0 configuration: update replication tutorials (resolving conflicts)

3d33b6d

3.0 configuration: update per TW review

33dd5af

3.0 configuration: update per TW review 2

1a0aa6f

andreyaksenov force-pushed the 3.0-config-replication-tutorials branch from b195ffe to 1a0aa6f Compare December 7, 2023 07:10

andreyaksenov merged commit 3eab90f into 3.0 Dec 7, 2023

andreyaksenov deleted the 3.0-config-replication-tutorials branch December 7, 2023 07:19

	Make sure that ``upstream.status`` and ``downstream.status`` are ``follow`` for ``instance002`` and ``instance003``.
	Make sure that ``upstream.status`` and ``downstream.status`` are ``follow`` for ``instance001`` and ``instance003``.

	To check that replicas (``instance001`` and ``instance003``) get all updates from the master(``instance002``), follow the steps below:
	To check that replicas (``instance001`` and ``instance003``) get all updates from the master (``instance002``), follow the steps below:


		3. Use the ``select`` operation on ``instance001`` and ``instance003`` to make sure data is replicated.

		4. Check that the 1-st component of :ref:`box.info.vclock <box_introspection-box_info>` values are the same on all instances:

	First, set the :ref:`replication.failover <configuration_reference_replication_failover>` option to ``manual``:
	Set the :ref:`replication.failover <configuration_reference_replication_failover>` option to ``manual``:

	For ``instance002``, ``upstream.status`` and ``downstream.status`` should be ``follow``.
	For ``instance001``, ``upstream.status`` and ``downstream.status`` should be ``follow``.


		.. _replication-automated-failover-tt-env:

		Prerequisites

Conversation

andreyaksenov commented Nov 17, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Totktonada left a comment

Choose a reason for hiding this comment

Uh oh!

p7nov left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andreyaksenov Dec 5, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

andreyaksenov commented Nov 17, 2023 •

edited

Loading

andreyaksenov Dec 5, 2023 •

edited

Loading