How to do Distributed Transactions the RIGHT way? Microservices

[@xellos5262]

Concise, simple visuals, clean and non-monotone speech, in short fabulous. Thank you for the video, rare to get a good overview in such short time!

[@Vaibhavatutube]

Brilliant explanation and presentation. Thanks!

[@ammabalasimha]

Explanation is very clear. Thanks a lot

[@andersonanjos4455]

Perfectly explained. Thanks a lot.

[@phillipdino]

Such a grea video. Thank you for the explanation.

[@AK-vx4dy]

Very clear explanation of realy complicated things

[@BitsAndBytes0101]

You explained it very well. Thank a lot

[@blasttrash]

nice video. but one question at 6:58, while doing backward recovery, what if one of the backward step also fails? 😅Then state of our system is stuck as we cant move forward nor can we move backward.

[@ashishmaurya3669]

That's why it preferred async communication between microservice so request doesn't lose the track. Also if we go for sync we need to implement failover mechanism for eg retry the request to certain number.

[@evoxtorm]

We can also use Try-Confirm/Cancel approach also but we have to take care of out-of-order execution. But yes Saga are preferred due to their liner order of execution

[@hariyadi4833]

It appears quite intriguing. What is the most effective method for handling events, such as those related to notifications? In the event of an error within each service, it is anticipated that we can also issue a notification to consumers. However, is it considered secure to implement this across all events in every service?

[@premraj.m]

excellent, this is the perfect way to teach, 13mins went off very fast for me while watching this video

[@mohammedhuzaifa8654]

thanks dd for putting it in a way that is easy to understand

[@rezafarhang7268]

Amazing! short and helpful

[@mhopado]

Thanks for explaining such complicated topics in simple way. In real world you, will find the hybrid approach, where a orchestrator service keep track of the state of the transaction while passing the events from one service to other. So in case of failure, the orchestrator will retry from the failure state of the transaction or revert back the previous operation.

[@qichengri5517]

COOL! Then your messaging/queuing system doesn't need to be as reliable.

[@AizazShahid-ck8cn]

Amazing video, it would be great if you could explain the failure recovery process in those two saga implementations

[@rakeshkhuntia2093]

Thank you so much, Its very simple and clear explanation.

[@Md_sadiq_Md]

Pushing the algorithm ❤️

[@warriorfb2010]

Excellent. I look forward to watching your other videos. Thank you!

[@ManivannanBgi]

Thank you so much. it is really great explanation for us. thank a lot

[@SaravananViswanathan-t2i]

precise and simple. that created excellent content

[@sreejeshraj4800]

Thank you very much for your videos - excellent content, to the point. I have a request - can you please do a video comparing the similarities / differences between various styles of programming - messaging, event driven, asynchronous, reactive etc. I know that you have done vides on messaging, pub sub etc. - but a comparison video will make these concepts clear rather than doing separate videos for each topic.

[@80-two]

Avoid distributed transactions at ALL costs in microservices - just use events and eventual consistency. This approach causes so many headaches I can't even list them all here

What if I need strong consistency?

[@jeevan19876]

Sagas are eventually consistent though they are distributed transactions. The events from one service to another are usually asynchronous. Two phase commit is used to achieve strong consistency snd should be avoided

[@80-two]

@@jeevan19876 - SAGA's are a serious complication that you need to be 100% sure you need before applying.. There are far simpler patterns that solve the problems just fine IMO

[@gauravmatta8783]

Agreed but many times we may require an orchestration as well specially in synchronous communication in that case this approach is helpfull.

[@rajchaurasiya1265]

Thank you mam for this valuable content 🙏

[@williamgomes3730]

Very good video. Thanks a lot!

[@kotojava]

how to sort issues when some process just can't have compensations? eg we send money or we send email with some good eg ebook. So these actions can't be undone.

[@jeevan19876]

There is a pivot point in transactions, if any failure occurs after pivot the services would retry the steps and run for completion Else if pivot is failed just rollback transactions before pivot

[@kotojava]

@@jeevan19876 some actions might not be rollbacked.

[@duffy9090]

Great video, thank you for putting these topics together

[@ashishjaiswal1090]

Nice explanation through animated slides seems do a lot of work thanks for explaining in intersting way, is there any source/link using java it will be helpful.

[@shashavalijava9]

I 💕💕💕 your explanation . thanks you

[@sachinshukla1095]

Beautiful explanation....❤

[@blossoms2u]

Thank you !!!

[@nagareddy9744]

thank you

[@tendaikawadza4492]

how do i add user details to each service

[@jeevan19876]

Your service would be communicating with user service to get details based on id Each kafka event would contain a userid

[@soni.himansh]

Great video

[@ravitejaakkala6114]

well explained !!

[@mr_code11100]

Thanks for thats explication.

[@rishiraj2548]

thanks a lot

[@prithvirajan105]

Thanks...

[@AkshayMukadam]

live demo would have been great i.e hands on would have been good

[@Markcarleous1903]

Wasted hours in learning about distributed transaction 2phase commit and 3phase commit and SAGA. Then I was wondering why don't people seem to use 2phase commit while making system design videos. Here in this 15mins videos everything is clarified. Kash phle yhi dekhi hoti

[@bhaskarsadawarte7696]

🙌

[@jeevan19876]

I am in disagreement with that we need to avoid distributed transactions, distributed transactions are very common in microservices and sagas are used most of the time to ensure eventual consistency.

[@rohigt5745]

2PC was the old SOA style approach. I've never come across using 2PC in modern microservices. You always use eventual consistency using replication, segregate read & write state databases, quorum based polling to reconcile and finalise transaction state (eventually). Its always compensating transactions given the scale of data & concurrency microservices are intended to handle. 2PC will not be practical.

[@chessmaster856]

There are more and more failue points. You just keep doing lot more work keep lot of logs. Try again and just do too much work. It us easier to talk and go into architecture and blame developers failures in implementation.. there are too many failure scenarios possible.the best you can do is have robust hardware that wo t fail and kee forensic record or log of every step before doing and be able to go back and analyze retry. All this is easier in chat sites and more risky in financial world. You Tube can easily say that your count may nit be accurate on so and so date. That is fine. That is not possible in financial transavtions

[@chessmaster856]

Most architects have never implemented this in real life and would avoid going near development

[@chessmaster856]

But this video highlights most of issues in distributed transaction correctly

[@s.sambit1676]

Mam, You start looking old😔 but ur explanation growing sharply.

[@sudocode]

Well dear ignorant youngster you should probably know that I am older than I look and younger than I feel. Sad that’s the most interesting thing you observed about me 😅

[@theunusual4566]

What a shit comment on the Excellent video. Mam you are amazing as always.

[@warriorfb2010]

And you're immature and likely not worthy of working within a professional business domain.

[@CommonSense-person]

​@@sudocode kya thoka usko wah

[@ShortGiant1]

What a completely unnecessary comment on a nice video