Apache Cassandra - Tutorial 6 - Partitioning, Rings and Tokens

[@Thundernixx]

This was one the most explainable videos that I've saw about this topic, thank you so much for sharing it.

[@jumpstartCS]

Really glad you found it useful! Thanks for the support!

[@ahmetcihan8025]

Keep it up man... Thats one for the books.

[@asefnak4764]

Thanks man day 3 into Cassandra and I've completed 6th today

[@jumpstartCS]

Great to hear, thanks for watching!

[@donrounds7791]

minor observation - should be referring to tokens in a range that includes either the lower or upper bound. If the number is between 10 and 20 then that means 10 and 20 wont be mapped to a node. So either >=10 n 10 n

[@kbsunil7464]

since Cassandra is highly available and fault tolerant .ie if any node fails the data should be available from other nodes. but in you example all partitions with bmw resides in node one. what happens when node 1 fails .....from where did we get data for the partition bmw

[@santu4bth]

sharding is distribution or partition of data across multiple different machines whereas partitioning is distribution of data on the same machine

[@valerylukin8446]

Other nodes will have copy of the data from other nodes. This is controlled by replication factor that is set when you create a Keyspace. For example: You have 5 nodes and created a Keyspace with replication factor = 2. Then you have emergency with one of the node. Then Reads that this nodes suppose to be serving will be served from another node, since replication factor was set to 2 and one node contains duplicate data from failed node. I am not in anyway advanced with Cassandra, but this is the way I understand it will behave, based on my knowledge.

[@guents]

So will a query across multiple partitions work but be inefficient, or it won't work at all?

[@rohitmania1]

Question - What happens when a node runs out of capacity? Is this practically possible? Say a bunch of tokens for any given node represent too much data and needs to be split across more than one nodes. How do we handle that?

[@heh2k]

You watch df and add servers (with more drives) when the nodes fill up. You need to add enough nodes to take a slice of keyspace (the token ranges) from each existing node.

[@HarshaVardhan-jf9sd]

If i have a partition key that is being used heavily, then the data that is stored in the node against that partition key would be more than other nodes, this introduces non uniform distribution of data with some nodes being heavily loaded and other not...how do we deal with such kind of issues?

[@jumpstartCS]

I would recommend rethinking your partitioning strategy. It is important to partition on an appropriate value and think long term to make sure this will not cause issues down the line.

[@muhammedsalama8323]

Thanks for the great videos, please do you have monitoring shell scripts for cassendra for crontab jobs

[@jumpstartCS]

Hello, thanks for the support! I don't have any scripts handy but I would recommend checking out the book Cassandra: The Definitive Guide by O'Reilly as it might have some content that will help you!

[@madhusudandad6028]

What if a node has exceeds it's storge limit based on token rane and new entry comes? Now I have setup cassandra on my machine...so it has only one node...so by default everything will go in the node only right?

[@jackcasey1502]

Great video! Anyone else think the ring diagram at 6:04 looks like a turtle?

[@amanali9501]

If i have a 100 different partition key ,so do i need 100 different machines?

[@jumpstartCS]

No you do not. A single node (machine) can handle one or more partitions. This is very common.

[@summary017]

How does 63 value decided ? 2 to the power of 63 .... Where is 63 configured and what does it refer to ?

[@amanali9501]

It is 64 bit number range , do Google it

[@abhinee]

do we need to worry about token assignment while creating tables in cassandra?

[@AwaraGhumakkad]

how do we decide partition key ? what is a best way to decide? just guessing - is it the common column between 2 tables?

[@smitalpatel]

not the common column but the column on which you will be searching with where clause. Reason is that particular key should be on same partition so that the retrieval is faster.

[@aswathis2938]

why cant i query my table in cassandra using primary key? It will work right then why its not according to your video?

[@wulymammoth]

Your question is unclear -- are you talking about query by PK on the table whose partition key is "car make" or the second table where the partition key is the "car ID" which is the same as our idea of a PK in a relational DB? If you're querying by PK, you want to build a table whose partition key is your PK...

[@aswathis2938]

@@wulymammoth I am a beginner in cassandra, My question is if suppose say i define primary key like PRIMARY KEY((car_id), car_make) then if i use select statement like SELECT * FROM CAR WHERE car_id = "some" and car_make = "some" is it required to mention both car_id & car_make in "where" clause in cassandra query or just "car_id"?

[@progfan234]

Thanks

[@jumpstartCS]

Thanks so much for your support!

[@progfan234]

@jumpstartCS you're welcome. Thanks once again!

[@heh2k]

But how are the tokens assigned? This seems to be totally undocumented, which is bizarre because it's critical to data safety.

[@abirpaul9027]

Thanks for sharing the knowledge

[@jumpstartCS]

No problem thanks for watching!

[@donl7737]

Why does it sound somewhat similar to consistent having technique?

[@melodiessim2570]

it is the consistent hashing (according to Cassandra's website)

[@cantwaittowatch]

Excellent!

[@Andrzej3935]

Thanks!

[@shivendrakadam7596]

Awesome

[@198273sam]

That's awesome

[@jumpstartCS]

Thanks for the support Samir!