For what reason should I use a real multi-Primary setup?
To be clear, not a multi-writer solution where any node can become the active writer in case of needs, as for PXC or PS-Group_replication.
No, we are talking about a multi-Primary setup where I can write at the same time on multiple nodes.
I want to insist on this “why?”.

After having excluded the possible solutions mentioned above, both covering the famous 99,995% availability, which is 26.30 minutes downtime in a year, what is left?

Disaster Recovery? Well that is something I would love to have, but to be a real DR solution we need to put several kilometers (miles for imperial) in the middle. 

And we know (see here and here) that aside some misleading advertising, we cannot have a tightly coupled cluster solution across geographical regions.

So, what is left? I may need more HA, ok that is a valid reason. Or I may need to scale the number of writes, ok that is a valid reason as well.
This means, at the end, that I am looking to a multi-Primary because:

• Scale writes (more nodes more writes)
  • Consistent reads (what I write on A must be visible on B)
• Gives me 0 (zero) downtime, or close to that (5 nines is a maximum downtime of 864 milliseconds per day!!)
• Allow me to shift the writer pointer at any time from A to B and vice versa, consistently.   

Now, keeping myself bound to the MySQL ecosystem, my natural choice would be MySQL NDB cluster.

But my (virtual) boss was at AWS re-invent and someone mentioned to him that Aurora Multi-Primary does what I was looking for.

This (long) article is my voyage in discovering if that is true or … not.

Given I am focused on the behaviour first, and NOT interested in absolute numbers to shock the audience with millions of QPS, I will use low level Aurora instances. And will perform tests from two EC2 in the same VPC/region of the nodes.

You can find the details about the tests on GitHub here Finally I will test:

• Connection speed
• Stale read
• Write single node for baseline
• Write on both node:
  • Scaling splitting the load by schema
  • Scaling same schema 

Tests results

Let us start to have some real fun. The first test is … 

Connection Speed

The purpose of this test is to evaluate the time taken in opening a new connection and time taken to close it. The action of open/close connection can be a very expensive operation especially if applications do not use a connection pool mechanism.

As we can see ProxySQL results to be the most efficient way to deal with opening connections, which was expected given the way it is designed to reuse open connections towards the backend. 

Different is the close connection operation in which ProxySQL seems to take a little bit longer.  As global observation we can say that using ProxySQL we have more consistent behaviour. Of course this test is a simplistic one, and we are not checking the scalability (from 1 to N connections) but it is good enough to give us the initial feeling. Specific connection tests will be the focus of the next blog on Aurora MM. 

Stale Reads

Aurora MultiPrimary use the same mechanism of the default Aurora to update the buffer pool:

Using the Page Cache update, just doing both ways. This means that the Buffer Pool of Node2 is updated with the modification performed in Node1 and vice versa.

To verify if an application would be really able to have consistent reads, I have run this test. This test is meant to measure if, and how many, stale reads we will have when writing on a node and reading from the other.

Amazon Aurora multi Primary has 2 consistency model:

As an interesting fact the result was that with the default consistency model (INSTANCE_RAW), we got 100% stale read.

Given that I focused on identifying the level of the cost that exists when using the other consistency model (REGIONAL_RAW) that allows an application to have consistent reads.

The results indicate an increase of the 44% in total execution time, and of the 95% (22 time slower) in write execution. 

It is interesting to note that the time taken is in some way predictable and consistent between the two consistency models. 

The graph below shows in yellow how long the application must wait to see the correct data on the reader node. While in blue is the amount of time the application waits to get back the same consistent read because it must wait for the commit on the writer.

As you can see the two are more or less aligned. Given the performance cost imposed by using REGIONAL_RAW,  all the other tests are done the defaut INSTANCE_RAW, unless explicitly stated.

Writing tests

All tests run in this section were done using sysbench-tpcc with the following settings:

sysbench ./tpcc.lua --mysql-host=<> --mysql-port=3306 --mysql-user=<> --mysql-password=<> --mysql-db=tpcc --time=300 --threads=32 --report-interval=1 --tables=10 --scale=15  --mysql_table_options=" CHARSET=utf8 COLLATE=utf8_bin"  --db-driver=mysql prepare

sysbench /opt/tools/sysbench-tpcc/tpcc.lua --mysql-host=$mysqlhost --mysql-port=$port --mysql-user=<> --mysql-password=<> --mysql-db=tpcc --db-driver=mysql --tables=10 --scale=15 --time=$time  --rand-type=zipfian --rand-zipfian-exp=0 --report-interval=1 --mysql-ignore-errors=all --histogram  --report_csv=yes --stats_format=csv --db-ps-mode=disable --threads=$threads run

Write Single node (Baseline)

Before starting the comparative analysis, I was looking to define what was the “limit” of traffic/load for this platform. 

From the graph above, we can see that this setup scales up to 128 threads after that, the performance remains more or less steady. 

Amazon claims that we can mainly double the performance when using both nodes in write mode and use a different schema to avoid conflict.

Once more remember I am not interested in the absolute numbers here, but I am expecting the same behaviour Given that our expectation is to see:

Write on both nodes different schemas

So AWS recommend this as the scaling solution:

And I diligently follow the advice.

I used 2 EC2 nodes in the same subnet of the Aurora Node, writing to a different schema (tpcc & tpcc2). 

Overview

Let us make it short and go straight to the point. Did we get the expected scalability?

Well no:

We just had 26% increase, quite far to be the expected 100% Let us see what happened in detail (if not interested just skip and go to the next test).

Node 1

Node 2

As you can see Node1 was (more or less) keeping up with the expectations and being close to the expected performance.
But Node2 was just not keeping up, performances there were just terrible. 
The graphs below show what happened.

While Node1 was (again more or less) scaling up to the baseline expectations (128 threads), Node2 collapsed on its knees at 16 threads. Node2 was never able to scale up.

Reads

Node 1

Node1 is scaling the reads as expected also if here and there we can see performance deterioration.

Node 2

Node2 is not scaling Reads at all. 

Writes

Node 1

Same as Read

Node 2

Same as read

Now someone may think I was making a mistake and I was writing on the same schema. I assure you I was not.

Check the next test to see what happened if using the same schema.  

Write on both nodes same schema

Overview

Now, now Marco, this is unfair. You know this will cause contention.

Yes I do! But nonetheless I was curious to see what was going to happen and how the platform would deal with that level of contention. 
My expectations were to have a lot of performance degradation and increased number of locks. About conflict I was not wrong, node2 after the test reported:

+-------------+---------+-------------------------+
| table       | index   | PHYSICAL_CONFLICTS_HIST |
+-------------+---------+-------------------------+
| district9   | PRIMARY |                    3450 |
| district6   | PRIMARY |                    3361 |
| district2   | PRIMARY |                    3356 |
| district8   | PRIMARY |                    3271 |
| district4   | PRIMARY |                    3237 |
| district10  | PRIMARY |                    3237 |
| district7   | PRIMARY |                    3237 |
| district3   | PRIMARY |                    3217 |
| district5   | PRIMARY |                    3156 |
| district1   | PRIMARY |                    3072 |
| warehouse2  | PRIMARY |                    1867 |
| warehouse10 | PRIMARY |                    1850 |
| warehouse6  | PRIMARY |                    1808 |
| warehouse5  | PRIMARY |                    1781 |
| warehouse3  | PRIMARY |                    1773 |
| warehouse9  | PRIMARY |                    1769 |
| warehouse4  | PRIMARY |                    1745 |
| warehouse7  | PRIMARY |                    1736 |
| warehouse1  | PRIMARY |                    1735 |
| warehouse8  | PRIMARY |                    1635 |
+-------------+---------+-------------------------+

Which is obviously a strong indication something was not working right. In terms of performance gain, if we compare ONLY the result with the 128 Threads :

Also with the high level of conflict we still have 12% of performance gain.

The problem is that in general we have the two nodes behave quite badly.
If you check the graph below you can see that the level of conflict is such to prevent the nodes not only to scale but to act consistently.

Node 1

Node 2

Reads

In the following graphs we can see how node1 had issues and it actually crashed 3 times, during tests with 32/64/512 treads.
Node2 was always up but the performances were very low. 

Node 1

Node 2

Writes

Node 1

Node 2

Recovery from crashed Node

About recovery time reading the AWS documentation and listening to presentations, I often heard that Aurora Multi Primary is a 0 downtime solution.
Or other statements like: “in applications where you can't afford even brief downtime for database write operations, a multi-master cluster can help to avoid an outage when a writer instance becomes unavailable. The multi-master cluster doesn't use the failover mechanism, because it doesn't need to promote another DB instance to have read/write capability”

To achieve this the suggestion I found, was to have applications pointing directly to the Nodes endpoint and not use the Cluster endpoint.
In this context the solution pointing to the Nodes should be able to failover within a seconds or so, while the cluster endpoint:

Personally I think that designing an architecture where the application is responsible for the connection to the database and failover is some kind of refuse from 2001. But if you feel this is the way, well go for it.

What I did for testing is to use ProxySQL, as plain as possible, with nothing else then the basic monitor coming from the native monitor.

I then compare the results with the tests using the Cluster endpoint.
In this way I adopt the advice of pointing directly at the nodes, but I was doing things in our time.  

The results are below and they confirm (more or less) the data coming from Amazon.

A downtime of 7 seconds is quite a long time nowadays, especially if I am targeting the 5 nines solution that I want to remember is 864 ms downtime per day.

Using ProxySQL is going closer to that, still too long to be called 0 (zero) downtime.

I also have fail-back issues when using the AWS cluster endpoint.

Given it was not able to move the connection to the joining node seamlessly. 

Last but not least when using the consistency level INSTANCE_RAW, I had some data issue as well as PK conflict:
FATAL: mysql_drv_query() returned error 1062 (Duplicate entry '18828082' for key 'PRIMARY')   

Conclusions

As state the beginning of this long blog the reasons expectations to go for a multi Primary solution were:

• Scale writes (more nodes more writes)
• Gives me 0 (zero) downtime, or close to that (5 nines is a maximum downtime of 864 milliseconds per day!!)
• Allow me to shift the writer pointer at any time from A to B and vice versa, consistently.   

Honestly I feel we have completely failed the scaling point.

Probably if I use the largest Aurora I will get much better absolute numbers, and it will take me more to encounter the same issues, but I will.

In any case if the Multi muster solution is designed to provide that scalability, it should do that with any version.

I did not have zero downtime, but I was able to failover pretty quickly with ProxySQL.

Finally, unless the consistency model is REGIONAL_RAW, shifting from one node to the other is not prone to possible negative effects like stale reads.

Because that I consider this requirement not satisfied in full. 

Given all the above, I think this solution could eventually be valid only for High Availability (close to be 5 nines), but given it comes with some limitations I do not feel comfortable in preferring it over others just for HA, at the end default Aurora is already good enough as a High available solution. 

references

https://www.youtube.com/watch?v=p0C0jakzYuc

https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/aurora-multi-master.html

https://www.slideshare.net/marcotusa/improving-enterprises-ha-and-disaster-recovery-solutions-reviewed

https://www.slideshare.net/marcotusa/robust-ha-solutions-with-proxysql

https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/aurora-multi-master.html#aurora-multi-master-limitations  

If you have a business no matter how small, you are collecting data, and you need to have your data accessible to make informed decisions about how to make your business better. The more successful you become the more data you are producing and the more you become dependent by it. This is when you start to realize your must have your data in a safe place like a database instead some spreadsheet.

But to put your data in a database is not enough, you must be sure the database will be robust, resilient, and always available when you or your customers need it.

When design architectures for robust database architectures, we always discuss about High Availability (HA) and Disaster Recovery (DR). These two concepts are elements of the larger umbrella that is the Business continuity plan.

To cover the different needs, we (Database Architects) use/apply two main models: Tightly Coupled cluster, and Loosely Coupled cluster (latest presentation https://www.slideshare.net/marcotusa/best-practicehigh-availabilitysolutiongeodistributedfinal).

The combination of the two different model allow us to build solid, resilient, scalable HA solution geographically distributed.

Like this:

Until now the part that was NOT natively supported in the architecture above was how to failover the replication channel from a node to another node in case of crash.

This element is currently cover by custom script(https://github.com/y-trudeau/Mysql-tools/tree/master/PXC) develop by my colleague Yves Trudeau.

But Oracle in MySQL 8.0.22 introduced the Asynchronous failover feature. In this blog I am going to check if this feature is enough to cover what we need to avoid using external movable/custom script or if instead there is still work to do.

My scenario is the following:

I have a PXC 8.0.20 cluster on DC1 and I am going to replicate towards a Group Replication cluster 8.0.22.

First, I need to establish the replication, to do so I follow the standard steps:

• Take a backup from one PXC node
• Copy over to DC2 and have a SINGLE MySQL 8.0.22 node start.
• Once all the internal upgrades are done, I am ready to create the Group Replication cluster as for https://dev.mysql.com/doc/refman/8.0/en/group-replication-deploying-in-single-primary-mode.html

Once I am up and running and have a situation like this:

[code]+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 38809616-e149-11ea-b0cc-080027d0603e | gr1         |        3306 |       ONLINE |     PRIMARY |         8.0.22 |
| group_replication_applier | 959cc074-e14b-11ea-b90c-080027d0603e | gr2         |        3306 |       ONLINE |   SECONDARY |         8.0.22 |
| group_replication_applier | a56b38ec-e14b-11ea-b9ce-080027d0603e | gr3         |        3306 |       ONLINE |   SECONDARY |         8.0.22 |
| group_replication_applier | b7039e3b-f744-11ea-854c-08002734ed50 | gr4         |        3306 |       ONLINE |   SECONDARY |         8.0.22 |
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+[/code]

Then I am ready to create the replication channel.

To do so I am first doing keeping the OLD style without Auto-failover. To do so:

On one of the PCX nodes:

Create user replica_dc@’192.168.4.%’ identified by ‘secret’;
Grant REPLICATION SLAVE on *.* to replica_dc@’192.168.4.%’;

On the PRIMARY of my Grou Replication cluster:

CHANGE MASTER to
master_user='<user>',
master_password='<secret>',
master_host='192.168.4.22',
master_auto_position=1,
source_connection_auto_failover=0,
master_retry_count=6,
master_connect_retry=10
for channel "dc1_to_dc2";

A brief explanation about the above:

• source_connection_auto_failover=0,  <--This enable/disable auto-failover
• master_retry_count=6,                       <-- The number of attempts
• master_connect_retry=10                  <-- Specifies the interval between reconnection attempts

Once Replication is up, we will have this scenario:

Now in case of a crash of the PXC node from which I replicate for, my Group replication cluster will stop replicating:

To prevent this to happen I will use the new functionality to allow Asynchronous replication failover.

To do so the first thing I have to do is to add the list of possible SOURCE in the new table that is locate in MySQL (mysql.replication_asynchronous_connection_failover), to do it I don’t need to use an insert command but instead a function:

asynchronous_connection_failover_add_source(channel, host, port, network_namespace, weight)

where:

• channel: The replication channel for which this replication source server is part of the source list.
• host: The host name for this replication source server.
• port: The port number for this replication source server.
• network_namespace: The network namespace for this replication source server (when specified).
• weight: The priority of this replication source server

In my case I will have:

SELECT asynchronous_connection_failover_add_source('dc1_to_dc2', '192.168.4.22', 3306, null, 100);
SELECT asynchronous_connection_failover_add_source('dc1_to_dc2', '192.168.4.23', 3306, null, 80);
SELECT asynchronous_connection_failover_add_source('dc1_to_dc2', '192.168.4.233', 3306, null, 50);

Now, and this is important, given the information is located in a table in the mysql.schema , the information present in the Primary is automatically replicated all over the cluster. Given that all nodes SEE the list of potential Source.

But given we are running our Group Replication cluster in Single Primary mode we can have only the Primary acting as REPLICA, because only the Primary can write. Hope that is clear.

To activate the failover we just need to pass the command:

First let us check the status:

SELECT CHANNEL_NAME, SOURCE_CONNECTION_AUTO_FAILOVER FROM performance_schema.replication_connection_configuration where CHANNEL_NAME = 'dc1_to_dc2';
+--------------+---------------------------------+
| CHANNEL_NAME | SOURCE_CONNECTION_AUTO_FAILOVER |
+--------------+---------------------------------+
| dc1_to_dc2   | 0                               |
+--------------+---------------------------------+

Ok all good I can activate the auto_failover:

stop slave for channel 'dc1_to_dc2';
CHANGE MASTER TO source_connection_auto_failover=1 for channel "dc1_to_dc2";
start slave for channel 'dc1_to_dc2';
SELECT CHANNEL_NAME, SOURCE_CONNECTION_AUTO_FAILOVER FROM performance_schema.replication_connection_configuration where CHANNEL_NAME = 'dc1_to_dc2';

+--------------+---------------------------------+
| CHANNEL_NAME | SOURCE_CONNECTION_AUTO_FAILOVER |
+--------------+---------------------------------+
| dc1_to_dc2   | 1                               |
+--------------+---------------------------------+

Perfect all is running, and this is my layout now:

At this point if something happens to my current SOURCE the auto-failover will be triggered, and I should see my GR Primary move to the other node.

After I kill the pxc_node2 the error log of the primary will present:

[ERROR] [MY-010584] [Repl] Slave I/O for channel 'dc1_to_dc2': error reconnecting to master This email address is being protected from spambots. You need JavaScript enabled to view it.:3306' - retry-time: 10 retries: 1 message: Can't connect to MySQL server on '192.168.4.23' (111), Error_code: MY-002003
...
[ERROR] [MY-010584] [Repl] Slave I/O for channel 'dc1_to_dc2': error reconnecting to master This email address is being protected from spambots. You need JavaScript enabled to view it.:3306' - retry-time: 10 retries: 6 message: Can't connect to MySQL server on '192.168.4.23' (111), Error_code: MY-002003

[System] [MY-010562] [Repl] Slave I/O thread for channel 'dc1_to_dc2': connected to master This email address is being protected from spambots. You need JavaScript enabled to view it.:3306',replication started in log 'FIRST' at position 53656167
[Warning] [MY-010549] [Repl] The master's UUID has changed, although this should not happen unless you have changed it manually. The old UUID was 28ae74e9-12c7-11eb-8d57-08002734ed50.

The Primary identify the SOURCE failed, try N time as asked waiting for each try T seconds. After that it try to change the master choosing it form the given list. If we use a different weight value, the node that will be elected will be the active one with the highest weight value.

Once a node is elected as SOURCE, the replication channel WILL NOT fail back also if a node with higher weight value will become available. The replication SOURCE change only if there is a failure in the communication between SOURCE-REPLICA pair.

This is what I have now:

Is this enough?

Well let us say that we are very close, but no this is not enough.

Why?

Because this cover ONLY a side of the picture. Which is the REPLICA node able to change the SOURCE. But if we are talking of a cluster like Group Replication, we need to have a mechanism that will allow the CLUSTER (and not only the single node) to fail-over.

What does it mean?

Here I mean that IF the Primary fails (remember only a Primary can become a REPLICA because only a Primary can write), another Primary will be elected by the cluster, and I would expect that if my previous Primary was replicating from a SOURCE, then the new one will do the same starting from the last valid applied position.

I am sure Oracle had already identified this as a possible issue, given to me this new feature sounds something done to make the architecture based on MySQL more resilient. As such the above looks like the most logic step forward when thinking to Tightly Coupled cluster like Group Replication.

Summary of commands

Set the source:

change master to master_user='replica_dc', master_password=<secret>, master_host='192.168.4.22', master_auto_position=1, source_connection_auto_failover=0,   master_retry_count=6, master_connect_retry=10 for channel "dc1_to_dc2";

Set the list:

SELECT asynchronous_connection_failover_add_source('dc1_to_dc2', '192.168.4.22', 3306, null, 100);
SELECT asynchronous_connection_failover_add_source('dc1_to_dc2', '192.168.4.23', 3306, null, 80);
SELECT asynchronous_connection_failover_add_source('dc1_to_dc2', '192.168.4.233', 3306, null, 50);

Delete an entry from the list

SELECT asynchronous_connection_failover_delete_source('dc1_to_dc2', '192.168.4.22', 3306, '');

Check if a replication channel is using automatic failover:

SELECT CHANNEL_NAME, SOURCE_CONNECTION_AUTO_FAILOVER FROM performance_schema.replication_connection_configuration where CHANNEL_NAME = 'dc1_to_dc2';

To change the SOURCE setting:

stop slave for channel 'dc1_to_dc2';
CHANGE MASTER TO source_connection_auto_failover=1 for channel "dc1_to_dc2";
start slave for channel 'dc1_to_dc2';

Conclusions

This is a very nice feature that will significantly increase the architecture resilience when using asynchronous replication in the architecture.

Still I think we need to wait to have it fully cover what is needed. This is a great first step but not enough, we still must manage the replication if a node in the REPLICA cluster fails.

I really hope we will see the evolution of that soon.

In the meantime, GREAT WORK MySQL/Oracle team, really!!

References

https://dev.mysql.com/doc/refman/8.0/en/replication-functions-source-list.html

https://www.slideshare.net/marcotusa/best-practicehigh-availabilitysolutiongeodistributedfinal

https://github.com/y-trudeau/Mysql-tools/tree/master/PXC

https://dev.mysql.com/doc/refman/8.0/en/change-master-to.html